row_id
string
section
string
section_slug
string
resource_type
string
marker
string
title
string
url
string
url_kind
string
domain
string
annotation
string
description
string
key_contribution
string
novelty
string
impact
string
signal
string
signal_strength
string
source_readme
string
source_line
int64
source_url
string
date_added
string
collection
string
collection_slug
string
user_goal
string
lifecycle_stages
string
audience
string
loop_layer
string
scope_fit
string
evidence_class
string
evidence_tier
string
source_status
string
canonical_url
string
source_title
string
source_description
string
authors
string
publication_date
string
publication_year
string
publication_venue
string
publisher
string
doi
string
publication_note
string
primary_category
string
metadata_source
string
github_repo
string
github_stars
string
github_forks
string
github_license
string
github_created_at
string
github_updated_at
string
arxiv_id
string
audited_at
timestamp[ms]
ale-0901
Examples And Schema
examples-and-schema
Template
🧾
Runnable test-repair loop
examples/runnable/test-repair-loop.sh
local_path
Repeats a failing deterministic check with durable progress, duplicate-failure detection, and a hard retry budget.
Repeats a failing deterministic check with durable progress, duplicate-failure detection, and a hard retry budget.
Repeats a failing deterministic check with durable progress, duplicate-failure detection, and a hard retry budget.
Durable execution and replay are treated as first-class loop infrastructure. Repeats a failing deterministic check with durable progress, duplicate-failure detection, and a hard retry budget.
Use Runnable test-repair loop to reuse a concrete artifact or connect it to the wider ecosystem.
Repository file; inspect the linked schema, example, guide, or implementation.
medium
README.md
1,634
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1634
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
verification;budget
builder;operator
workflow
direct
repository-native
A
local_ok
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/examples/runnable/test-repair-loop.sh
Runnable test-repair loop
2026
GitHub
GitHub
repository
2026-09-04T05:23:38
ale-0902
Examples And Schema
examples-and-schema
Template
🧾
Runnable loop guide
examples/runnable/README.md
local_path
Compares 8 starters by trigger, state, gate, and runtime, including executable test-repair, threshold-monitor, and queue-worker loops.
Compares 8 starters by trigger, state, gate, and runtime, including executable test-repair, threshold-monitor, and queue-worker loops.
Compares 8 starters by trigger, state, gate, and runtime, including executable test-repair, threshold-monitor, and queue-worker loops.
State persistence is explicit enough for repeated runs and handoff. Compares 8 starters by trigger, state, gate, and runtime, including executable test-repair, threshold-monitor, and queue-worker loops.
Use Runnable loop guide to reuse a concrete artifact or connect it to the wider ecosystem.
Repository file; inspect the linked schema, example, guide, or implementation.
medium
README.md
1,635
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1635
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
trigger;intake;verification;state
builder;operator
workflow
direct
repository-native
A
local_ok
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/examples/runnable/README.md
Runnable loop guide
2026
GitHub
GitHub
repository
2026-09-04T05:23:38
ale-0903
Community Gallery
community-gallery
Template
🧾
Loop gallery guide
gallery/README.md
local_path
Quality bar for contributed loop examples with receipts and lessons learned.
Quality bar for contributed loop examples with receipts and lessons learned.
Quality bar for contributed loop examples with receipts and lessons learned.
The resource is directly reusable as a starting artifact. Quality bar for contributed loop examples with receipts and lessons learned.
Use Loop gallery guide to reuse a concrete artifact or connect it to the wider ecosystem.
Repository file; inspect the linked schema, example, guide, or implementation.
medium
README.md
1,651
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1651
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
state
builder;operator
operations
direct
repository-native
A
local_ok
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/gallery/README.md
Loop gallery guide
2026
GitHub
GitHub
repository
2026-09-04T05:23:38
ale-0904
Community Gallery
community-gallery
Template
🧾
Loop gallery template
gallery/template.md
local_path
Markdown template for sharing a loop's trigger, intake, state, verification, escalation, and safety notes.
Markdown template for sharing a loop's trigger, intake, state, verification, escalation, and safety notes.
Markdown template for sharing a loop's trigger, intake, state, verification, escalation, and safety notes.
Verification is promoted from a final check to a loop-control signal. Markdown template for sharing a loop's trigger, intake, state, verification, escalation, and safety notes.
Use Loop gallery template to reuse a concrete artifact or connect it to the wider ecosystem.
Repository file; inspect the linked schema, example, guide, or implementation.
medium
README.md
1,652
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1652
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
trigger;intake;verification;state;escalation
builder;operator
operations
direct
repository-native
A
local_ok
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/gallery/template.md
Loop gallery template
2026
GitHub
GitHub
repository
2026-09-04T05:23:38
ale-0905
Community Gallery
community-gallery
Pattern
🔁
PR babysitter reference loop
gallery/pr-babysitter-reference.md
local_path
Reference gallery entry for keeping a pull request moving.
Reference gallery entry for keeping a pull request moving.
Reference gallery entry for keeping a pull request moving.
Turns loop adoption into shareable cases with enough structure to compare lessons learned. Reference gallery entry for keeping a pull request moving.
Use PR babysitter reference loop to reuse a concrete artifact or connect it to the wider ecosystem.
Repository file; inspect the linked schema, example, guide, or implementation.
medium
README.md
1,653
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1653
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder;operator
operations
direct
repository-native
A
local_ok
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/gallery/pr-babysitter-reference.md
PR babysitter reference loop
2026
GitHub
GitHub
repository
2026-09-04T05:23:38
ale-0906
Community Gallery
community-gallery
Pattern
🔁
CI repair reference loop
gallery/ci-repair-reference.md
local_path
Reference gallery entry for turning failing CI into a verified patch or escalation.
Reference gallery entry for turning failing CI into a verified patch or escalation.
Reference gallery entry for turning failing CI into a verified patch or escalation.
Verification is promoted from a final check to a loop-control signal. Reference gallery entry for turning failing CI into a verified patch or escalation.
Use CI repair reference loop to reuse a concrete artifact or connect it to the wider ecosystem.
Repository file; inspect the linked schema, example, guide, or implementation.
medium
README.md
1,654
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1654
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
verification;escalation
builder;operator
operations
direct
repository-native
A
local_ok
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/gallery/ci-repair-reference.md
CI repair reference loop
2026
GitHub
GitHub
repository
2026-09-04T05:23:38
ale-0907
Community Gallery
community-gallery
Pattern
🔁
Docs drift reference loop
gallery/docs-drift-reference.md
local_path
Reference gallery entry for recurring docs/code consistency checks.
Reference gallery entry for recurring docs/code consistency checks.
Reference gallery entry for recurring docs/code consistency checks.
Turns loop adoption into shareable cases with enough structure to compare lessons learned. Reference gallery entry for recurring docs/code consistency checks.
Use Docs drift reference loop to reuse a concrete artifact or connect it to the wider ecosystem.
Repository file; inspect the linked schema, example, guide, or implementation.
medium
README.md
1,655
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1655
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder;operator
operations
direct
repository-native
A
local_ok
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/gallery/docs-drift-reference.md
Docs drift reference loop
2026
GitHub
GitHub
repository
2026-09-04T05:23:38
ale-0908
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Critique
⚠️
Most Developers Do Not Need Agent Loops Yet
https://alphasignalai.substack.com/p/most-developers-do-not-need-agent
external
alphasignalai.substack.com
Useful caution against adopting loops before the task, signal, and economics justify them.
Useful caution against adopting loops before the task, signal, and economics justify them.
Useful caution against adopting loops before the task, signal, and economics justify them.
Keeps adoption grounded in known failure modes, economics, and operational limits. Useful caution against adopting loops before the task, signal, and economics justify them.
Use Most Developers Do Not Need Agent Loops Yet to bound risk before recurring or unattended execution.
Contextual source from alphasignalai.substack.com; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,663
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1663
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
operator;security
cross-layer
enabling
risk-analysis
B
ok
https://alphasignalai.substack.com/p/most-developers-do-not-need-agent
Most Developers Do Not Need Agent Loops Yet
The patterns were documented in 2024. Here’s who it pays off for, and the four conditions that decide.
AlphaSignal AI
Substack
html-meta
2026-09-04T05:23:38
ale-0909
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Critique
⚠️
Engineering Agentic Systems for Reliability
https://pruningmypothos.com/systems/engineering-agentic-systems-for-reliability/
external
pruningmypothos.com
Cautions that agentic systems fail at boundaries when permissions, verification, traceability, and escalation are weak.
Cautions that agentic systems fail at boundaries when permissions, verification, traceability, and escalation are weak.
Cautions that agentic systems fail at boundaries when permissions, verification, traceability, and escalation are weak.
Verification is promoted from a final check to a loop-control signal. Cautions that agentic systems fail at boundaries when permissions, verification, traceability, and escalation are weak.
Use Engineering Agentic Systems for Reliability to bound risk before recurring or unattended execution.
Contextual source from pruningmypothos.com; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,664
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1664
Govern
govern
Bound permissions, cost, failure, and escalation.
workspace;verification;escalation
operator;security
cross-layer
enabling
risk-analysis
B
ok
https://pruningmypothos.com/systems/engineering-agentic-systems-for-reliability/
Engineering Agentic Systems for Reliability | Sans Serif Systems
A practical reliability model for agentic systems built around governed steps, verification, escalation, and observability.
Shailesh Rawat
Sans Serif Systems
html-meta
2026-09-04T05:23:38
ale-0910
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Critique
⚠️
Self-Correcting Agents: Reflexion, CRITIC, and ReAct Loops Compared
https://callsphere.ai/blog/self-correcting-agents-reflexion-critic-react-loops-compared-2026
external
callsphere.ai
Compares self-correction patterns and their cost/failure tradeoffs.
Compares self-correction patterns and their cost/failure tradeoffs.
Compares self-correction patterns and their cost/failure tradeoffs.
Keeps adoption grounded in known failure modes, economics, and operational limits. Compares self-correction patterns and their cost/failure tradeoffs.
Use Self-Correcting Agents: Reflexion, CRITIC, and ReAct Loops Compared to bound risk before recurring or unattended execution.
Contextual source from callsphere.ai; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,665
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1665
Govern
govern
Bound permissions, cost, failure, and escalation.
verification;budget
operator;security
cross-layer
enabling
risk-analysis
B
restricted
https://callsphere.ai/blog/self-correcting-agents-reflexion-critic-react-loops-compared-2026
callsphere.ai
domain-fallback
2026-09-04T05:23:38
ale-0911
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Critique
⚠️
How to Build an AI Agent Harness: A 2026 Complete Guide
https://atlan.com/know/how-to-build-ai-agent-harness/
external
atlan.com
Broad guide with useful warnings on data readiness, permissions, context management, and evaluation.
Broad guide with useful warnings on data readiness, permissions, context management, and evaluation.
Broad guide with useful warnings on data readiness, permissions, context management, and evaluation.
Evaluation data is used as the feedback signal for improving loop behavior. Broad guide with useful warnings on data readiness, permissions, context management, and evaluation.
Use How to Build an AI Agent Harness: A 2026 Complete Guide to bound risk before recurring or unattended execution.
Contextual source from atlan.com; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,666
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1666
Govern
govern
Bound permissions, cost, failure, and escalation.
workspace;context;verification
operator;security
cross-layer
enabling
risk-analysis
B
ok
https://atlan.com/know/how-to-build-ai-agent-harness/
How to Build an AI Agent Harness: Step-by-Step Tutorial (2026)
Most agent harnesses fail at the data layer, not the loop. Build one the right way in 10 steps, with code and a done test for each. Start at Step 0.
atlan.com
domain-fallback
2026-09-04T05:23:38
ale-0912
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Critique
⚠️
Harness Engineering vs Prompt Engineering vs Context Engineering Explained
https://medium.com/@visrow/harness-engineering-vs-prompt-engineering-vs-context-engineering-explained-0423b692c87d
external
medium.com
Adjacent framing that helps avoid confusing loop engineering with the surrounding harness discipline.
Adjacent framing that helps avoid confusing loop engineering with the surrounding harness discipline.
Adjacent framing that helps avoid confusing loop engineering with the surrounding harness discipline.
Context is managed as durable loop state rather than a single prompt payload. Adjacent framing that helps avoid confusing loop engineering with the surrounding harness discipline.
Use Harness Engineering vs Prompt Engineering vs Context Engineering Explained to bound risk before recurring or unattended execution.
Contextual source from medium.com; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,667
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1667
Govern
govern
Bound permissions, cost, failure, and escalation.
context
operator;security
cross-layer
enabling
risk-analysis
B
restricted
https://medium.com/@visrow/harness-engineering-vs-prompt-engineering-vs-context-engineering-explained-0423b692c87d
Medium
domain-fallback
2026-09-04T05:23:38
ale-0913
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Position: Coding Benchmarks Are Misaligned with Agentic Software Engineering
https://arxiv.org/abs/2606.17799
external
arxiv.org
Argues benchmark scores conflate the model with the harness and penalize valid alternatives, so headline numbers hide which loop and harness choices actually move performance.
Argues benchmark scores conflate the model with the harness and penalize valid alternatives, so headline numbers hide which loop and harness choices actually move performance.
Argues benchmark scores conflate the model with the harness and penalize valid alternatives, so headline numbers hide which loop and harness choices actually move performance.
The work turns loop quality into a measurable task or score. Argues benchmark scores conflate the model with the harness and penalize valid alternatives, so headline numbers hide which loop and harness choices actually move performance.
Use Position: Coding Benchmarks Are Misaligned with Agentic Software Engineering to bound risk before recurring or unattended execution.
Research source arXiv:2606.17799; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,668
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1668
Govern
govern
Bound permissions, cost, failure, and escalation.
verification
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2606.17799
[2606.17799] Position: Coding Benchmarks Are Misaligned with Agentic Software Engineering
Coding agents have become a major mode of software engineering, but the benchmarks we use to compare them were designed in a pre-agent era: they collapse model, harness, and environment into a single end-to-end score, typically computed against one reference solution, with no component-level signal for iteration. We ar...
Maria I. Gorinova; Macey Baker; Amy Heineike; Maksim Shaposhnikov; Rob Willoughby; Dru Knox
2026-06-16
2026
arXiv
arXiv
cs.SE
arxiv-api
2606.17799
2026-09-04T05:23:38
ale-0914
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Understanding the Challenges in Iterative Generative Optimization with LLMs
https://arxiv.org/abs/2603.23994
external
arxiv.org
Empirically isolates three hidden design choices that make self-improving agent loops succeed or fail - starting artifacts, credit horizons over execution traces, and batching strategy - explaining why iterative refinement loops stay brittle in production.
Empirically isolates three hidden design choices that make self-improving agent loops succeed or fail - starting artifacts, credit horizons over execution traces, and batching strategy - explaining why iterative refinement loops stay brittle in production.
Empirically isolates three hidden design choices that make self-improving agent loops succeed or fail - starting artifacts, credit horizons over execution traces, and batching strategy - explaining why iterative refinement loops stay brittle in production.
Keeps adoption grounded in known failure modes, economics, and operational limits. Empirically isolates three hidden design choices that make self-improving agent loops succeed or fail - starting artifacts, credit horizons over execution traces, and batching strategy - explaining why iterative refinement loops stay bri...
Use Understanding the Challenges in Iterative Generative Optimization with LLMs to bound risk before recurring or unattended execution.
Research source arXiv:2603.23994; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,669
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1669
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2603.23994
[2603.23994] Understanding the Challenges in Iterative Generative Optimization with LLMs
Generative optimization uses large language models (LLMs) to iteratively improve artifacts (such as code, workflows or prompts) using execution feedback. It is a promising approach to building self-improving agents, yet in practice remains brittle: despite active research, only 9% of surveyed agents used any automated ...
Allen Nie; Xavier Daull; Zhiyi Kuang; Abhinav Akkiraju; Anish Chaudhuri; Max Piasevoli; Ryan Rong; YuCheng Yuan; Prerit Choudhary; Shannon Xiao; Rasool Fakoor; Adith Swaminathan; Ching-An Cheng
2026-03-25
2026
arXiv
arXiv
39 pages, 17 figures
cs.LG
arxiv-api
2603.23994
2026-09-04T05:23:38
ale-0915
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
The Illusion of Multi-Agent Advantage
https://arxiv.org/abs/2606.13003
external
arxiv.org
Systematic evaluation showing automatically generated multi-agent systems consistently underperform chain-of-thought self-consistency while costing up to 10x more, cautioning that auto-designed orchestration adds complexity without functional benefit.
Systematic evaluation showing automatically generated multi-agent systems consistently underperform chain-of-thought self-consistency while costing up to 10x more, cautioning that auto-designed orchestration adds complexity without functional benefit.
Systematic evaluation showing automatically generated multi-agent systems consistently underperform chain-of-thought self-consistency while costing up to 10x more, cautioning that auto-designed orchestration adds complexity without functional benefit.
Evaluation data is used as the feedback signal for improving loop behavior. Systematic evaluation showing automatically generated multi-agent systems consistently underperform chain-of-thought self-consistency while costing up to 10x more, cautioning that auto-designed orchestration adds complexity without functional b...
Use The Illusion of Multi-Agent Advantage to bound risk before recurring or unattended execution.
Research source arXiv:2606.13003; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,670
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1670
Govern
govern
Bound permissions, cost, failure, and escalation.
delegation;verification
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2606.13003
[2606.13003] The Illusion of Multi-Agent Advantage
Prevailing wisdom posits that Multi-Agent Systems (MAS) are superior to Single-Agent Systems (SAS), citing advantages like context protection, parallel processing and distributed decision-making. However, empirical support for this claim relies primarily on comparisons with SAS baselines using benchmarks that prioritiz...
Prathyusha Jwalapuram; Hehai Lin; Chuyuan Li; Fangkai Jiao; Sudong Wang; Yifei Ming; Zixuan Ke; Chengwei Qin; Giuseppe Carenini; Shafiq Joty
2026-06-11
2026
arXiv
arXiv
cs.AI
arxiv-api
2606.13003
2026-09-04T05:23:38
ale-0916
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Critique
⚠️
The Coming Loop
https://lucumr.pocoo.org/2026/6/23/the-coming-loop/
external
lucumr.pocoo.org
Flask creator Armin Ronacher's skeptical essay on harness loops, examining what continuously re-driving agents past their natural stopping points does to code quality, review capacity, and human understanding of the resulting systems.
Flask creator Armin Ronacher's skeptical essay on harness loops, examining what continuously re-driving agents past their natural stopping points does to code quality, review capacity, and human understanding of the resulting systems.
Flask creator Armin Ronacher's skeptical essay on harness loops, examining what continuously re-driving agents past their natural stopping points does to code quality, review capacity, and human understanding of the resulting systems.
Keeps adoption grounded in known failure modes, economics, and operational limits. Flask creator Armin Ronacher's skeptical essay on harness loops, examining what continuously re-driving agents past their natural stopping points does to code quality, review capacity, and human understanding of the resulting systems.
Use The Coming Loop to bound risk before recurring or unattended execution.
Contextual source from lucumr.pocoo.org; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,671
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1671
Govern
govern
Bound permissions, cost, failure, and escalation.
escalation;exit
operator;security
cross-layer
enabling
risk-analysis
B
ok
https://lucumr.pocoo.org/2026/6/23/the-coming-loop/
The Coming Loop | Armin Ronacher's Thoughts and Writings
Loops, harnesses, and why even loop skeptics may end up with them.
2026-06-23
2026
Armin Ronacher's Thoughts and Writings
html-meta
2026-09-04T05:23:38
ale-0917
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Critique
⚠️
Loop Engineering, the Latest AI Buzzword, Still Needs Humans in the Loop
https://www.theregister.com/ai-and-ml/2026/06/24/loop-engineering-latest-ai-buzzword-still-needs-humans-in-the-loop/5261735
external
www.theregister.com
The Register's report on the June 2026 loop-engineering discussion, collecting the Steinberger, Osmani, and Cherny quotes while arguing that vendor token-consumption incentives and model non-determinism keep humans in the loop.
The Register's report on the June 2026 loop-engineering discussion, collecting the Steinberger, Osmani, and Cherny quotes while arguing that vendor token-consumption incentives and model non-determinism keep humans in the loop.
The Register's report on the June 2026 loop-engineering discussion, collecting the Steinberger, Osmani, and Cherny quotes while arguing that vendor token-consumption incentives and model non-determinism keep humans in the loop.
Keeps adoption grounded in known failure modes, economics, and operational limits. The Register's report on the June 2026 loop-engineering discussion, collecting the Steinberger, Osmani, and Cherny quotes while arguing that vendor token-consumption incentives and model non-determinism keep humans in the loop.
Use Loop Engineering, the Latest AI Buzzword, Still Needs Humans in the Loop to bound risk before recurring or unattended execution.
Contextual source from www.theregister.com; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,672
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1672
Govern
govern
Bound permissions, cost, failure, and escalation.
budget
operator;security
cross-layer
enabling
risk-analysis
B
ok
https://www.theregister.com/ai-and-ml/2026/06/24/loop-engineering-latest-ai-buzzword-still-needs-humans-in-the-loop/5261735
Loop engineering, latest AI buzzword, still needs humans in the loop
Prompting less and automating more comes with a price
2026-06-24
2026
theregister
html-meta
2026-09-04T05:23:38
ale-0918
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
When Agents Do Not Stop: Uncovering Infinite Agentic Loops in LLM Agents
https://arxiv.org/abs/2607.01641
external
arxiv.org
Characterizes infinite agentic loops, a failure class where unbounded feedback paths make agents repeat calls, tools, or handoffs forever, and ships IAL-Scan, a static analyzer that confirmed 68 real cases across 47 of 6,549 scanned agent projects at 91.9% precision.
Characterizes infinite agentic loops, a failure class where unbounded feedback paths make agents repeat calls, tools, or handoffs forever, and ships IAL-Scan, a static analyzer that confirmed 68 real cases across 47 of 6,549 scanned agent projects at 91.9% precision.
Characterizes infinite agentic loops, a failure class where unbounded feedback paths make agents repeat calls, tools, or handoffs forever, and ships IAL-Scan, a static analyzer that confirmed 68 real cases across 47 of 6,549 scanned agent projects at 91.9% precision.
Keeps adoption grounded in known failure modes, economics, and operational limits. Characterizes infinite agentic loops, a failure class where unbounded feedback paths make agents repeat calls, tools, or handoffs forever, and ships IAL-Scan, a static analyzer that confirmed 68 real cases across 47 of 6,549 scanned agen...
Use When Agents Do Not Stop: Uncovering Infinite Agentic Loops in LLM Agents to bound risk before recurring or unattended execution.
Research source arXiv:2607.01641; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,673
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1673
Govern
govern
Bound permissions, cost, failure, and escalation.
workspace;delegation;exit
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.01641
[2607.01641] When Agents Do Not Stop: Uncovering Infinite Agentic Loops in LLM Agents
LLM agents increasingly rely on iterative execution to solve tasks through planning, tool use, state updates, and agent collaboration. While this design enables flexible automation, it also creates a new class of failures: an agent may repeatedly execute model calls, tools, workflow transitions, or agent handoffs when ...
Xinyi Hou; Shenao Wang; Yanjie Zhao; Haoyu Wang
2026-07-02
2026
arXiv
arXiv
cs.SE
arxiv-api
2607.01641
2026-09-04T05:23:38
ale-0919
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
The Blind Curator: How a Biased Judge Silently Disables Skill Retirement in Self-Evolving Agents
https://arxiv.org/abs/2607.07436
external
arxiv.org
Shows via corrupted-reward analysis that false-pass bias in an LLM judge silently disables the skill-retirement mechanism that keeps a self-evolving agent's growing skill library from drifting below the no-skill baseline.
Shows via corrupted-reward analysis that false-pass bias in an LLM judge silently disables the skill-retirement mechanism that keeps a self-evolving agent's growing skill library from drifting below the no-skill baseline.
Shows via corrupted-reward analysis that false-pass bias in an LLM judge silently disables the skill-retirement mechanism that keeps a self-evolving agent's growing skill library from drifting below the no-skill baseline.
Keeps adoption grounded in known failure modes, economics, and operational limits. Shows via corrupted-reward analysis that false-pass bias in an LLM judge silently disables the skill-retirement mechanism that keeps a self-evolving agent's growing skill library from drifting below the no-skill baseline.
Use The Blind Curator: How a Biased Judge Silently Disables Skill Retirement in Self-Evolving Agents to bound risk before recurring or unattended execution.
Research source arXiv:2607.07436; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,674
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1674
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.07436
[2607.07436] The Blind Curator: How a Biased Judge Silently Disables Skill Retirement in Self-Evolving Agents
A self-evolving agent retires its bad skills by watching them fail, so what happens when the judge cannot see the failures? Skill retirement is the structural constraint that keeps a growing library from drifting below the no-skill baseline, but its guarantee assumes an unbiased reward, which is false for the LLM judge...
Xing Zhang; Yanwei Cui; Guanghui Wang; Ziyuan Li; Wei Qiu; Bing Zhu; Peiyang He
2026-07-08
2026
arXiv
arXiv
Published at COLM 2026 Workshop on Agent Behavior
cs.AI
arxiv-api
2607.07436
2026-09-04T05:23:38
ale-0920
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Do LLM-Generated Skills Make Better AI Data Scientists? A Component Ablation Across Data-Science Workflows
https://arxiv.org/abs/2607.07504
external
arxiv.org
Negative result for low-curation skill libraries: across four data-science lifecycle stages (56 tasks), fully LLM-generated skill files show no reliable improvement over plain task prompting, and component ablations find no skill part that helps either (all p > 0.396). Useful counterweight to the skill-generation enthu...
Negative result for low-curation skill libraries: across four data-science lifecycle stages (56 tasks), fully LLM-generated skill files show no reliable improvement over plain task prompting, and component ablations find no skill part that helps either (all p > 0.396). Useful counterweight to the skill-generation enthu...
Negative result for low-curation skill libraries: across four data-science lifecycle stages (56 tasks), fully LLM-generated skill files show no reliable improvement over plain task prompting, and component ablations find no skill part that helps either (all p > 0.396). Useful counterweight to the skill-generation enthu...
Keeps adoption grounded in known failure modes, economics, and operational limits. Negative result for low-curation skill libraries: across four data-science lifecycle stages (56 tasks), fully LLM-generated skill files show no reliable improvement over plain task prompting, and component ablations find no skill part th...
Use Do LLM-Generated Skills Make Better AI Data Scientists? A Component Ablation Across Data-Science Workflows to bound risk before recurring or unattended execution.
Research source arXiv:2607.07504; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,675
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1675
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
researcher;evaluator;operator;security
cross-layer
enabling
research-paper
A
ok
https://arxiv.org/abs/2607.07504
[2607.07504] Do LLM-Generated Skills Make Better AI Data Scientists? A Component Ablation Across Data-Science Workflows
Product data scientists often ask LLM-based agents to help with recurring execution tasks such as cleaning data, writing SQL, choosing statistical tests, and formatting results. Reusable skill files are meant to avoid prompting from scratch by packaging guidance for a task family. Expert-written skills can encode high-...
Wei-Jung Huang
2026
2026
KDD Workshop on AI Data Scientist
ACM SIGKDD
Accepted at KDD Workshop on AI Data Scientist; the linked arXiv record is the available paper version.
cs.AI
Current arXiv acceptance note and official workshop page
2607.07504
2026-09-04T05:23:38
ale-0921
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
The Verification Horizon: No Silver Bullet for Coding Agent Rewards
https://arxiv.org/abs/2606.26300
external
arxiv.org
Position paper arguing verification has become harder than generation for coding agents: every verifier is only a proxy for underspecified human intent, so reward design faces a horizon that no single verification mechanism crosses.
Position paper arguing verification has become harder than generation for coding agents: every verifier is only a proxy for underspecified human intent, so reward design faces a horizon that no single verification mechanism crosses.
Position paper arguing verification has become harder than generation for coding agents: every verifier is only a proxy for underspecified human intent, so reward design faces a horizon that no single verification mechanism crosses.
Verification is promoted from a final check to a loop-control signal. Position paper arguing verification has become harder than generation for coding agents: every verifier is only a proxy for underspecified human intent, so reward design faces a horizon that no single verification mechanism crosses.
Use The Verification Horizon: No Silver Bullet for Coding Agent Rewards to bound risk before recurring or unattended execution.
Research source arXiv:2606.26300; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,676
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1676
Govern
govern
Bound permissions, cost, failure, and escalation.
verification;escalation
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2606.26300
[2606.26300] The Verification Horizon: No Silver Bullet for Coding Agent Rewards
A classical intuition holds that verifying a solution is easier than producing one. For today's coding agents, this intuition is being inverted: as foundation models develop stronger reasoning capabilities and engineering harnesses grow more sophisticated, generating complex candidate solutions is no longer difficult -...
Binghai Wang; Chenlong Zhang; Dayiheng Liu; Jiajun Zhang; Jiawei Chen; Mingze Li; Mouxiang Chen; Rongyao Fang; Siyuan Zhang; Xuwu Wang; Yuheng Jing; Zeyao Ma; Zeyu Cui
2026-06-24
2026
arXiv
arXiv
Authors are listed alphabetically by their first names
cs.AI
arxiv-api
2606.26300
2026-09-04T05:23:38
ale-0922
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Critique
⚠️
Write Code Like a Human Will Maintain It
https://unstack.io/write-code-like-a-human-will-maintain-it
external
unstack.io
Argues that agent-driven codebases create a compounding feedback loop where every merged shortcut becomes training signal for the next generation of changes, so code quality standards matter more, not less, under automation.
Argues that agent-driven codebases create a compounding feedback loop where every merged shortcut becomes training signal for the next generation of changes, so code quality standards matter more, not less, under automation.
Argues that agent-driven codebases create a compounding feedback loop where every merged shortcut becomes training signal for the next generation of changes, so code quality standards matter more, not less, under automation.
Keeps adoption grounded in known failure modes, economics, and operational limits. Argues that agent-driven codebases create a compounding feedback loop where every merged shortcut becomes training signal for the next generation of changes, so code quality standards matter more, not less, under automation.
Use Write Code Like a Human Will Maintain It to bound risk before recurring or unattended execution.
Contextual source from unstack.io; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,677
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1677
Govern
govern
Bound permissions, cost, failure, and escalation.
escalation
operator;security
cross-layer
enabling
risk-analysis
B
ok
https://unstack.io/write-code-like-a-human-will-maintain-it
Write code like a human will maintain it
One of the best things about LLMs is that they'll write code for you, all day long. Who cares about DRY? You don't have to be the one updating the same long con...
2026-07-10
2026
Unstack
html-meta
2026-09-04T05:23:38
ale-0923
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Rethinking the Evaluation of Harness Evolution for Agents
https://arxiv.org/abs/2607.12227
external
arxiv.org
Re-evaluates harness evolution on Terminal-Bench 2.1 with GPT-5.4 and Claude Opus 4.6, finding that evolved harnesses do not consistently beat budget-matched search and transfer only weakly to held-out tasks.
Re-evaluates harness evolution on Terminal-Bench 2.1 with GPT-5.4 and Claude Opus 4.6, finding that evolved harnesses do not consistently beat budget-matched search and transfer only weakly to held-out tasks.
Re-evaluates harness evolution on Terminal-Bench 2.1 with GPT-5.4 and Claude Opus 4.6, finding that evolved harnesses do not consistently beat budget-matched search and transfer only weakly to held-out tasks.
Evaluation data is used as the feedback signal for improving loop behavior. Re-evaluates harness evolution on Terminal-Bench 2.1 with GPT-5.4 and Claude Opus 4.6, finding that evolved harnesses do not consistently beat budget-matched search and transfer only weakly to held-out tasks.
Use Rethinking the Evaluation of Harness Evolution for Agents to bound risk before recurring or unattended execution.
Research source arXiv:2607.12227; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,678
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1678
2026-07-17
Govern
govern
Bound permissions, cost, failure, and escalation.
verification;budget
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.12227
[2607.12227] Rethinking the Evaluation of Harness Evolution for Agents
We revisit the evaluation of automatic harness evolution for LLM agents. Existing harness evolution methods use unit test cases to search for harness configurations and then report final performance on the same public benchmark. This protocol raises two fundamental concerns. First, harness evolution is itself an iterat...
Yike Wang; Huaisheng Zhu; Zhengyu Hu; Yige Yuan; Zhengyu Chen; Shakti Senthil; Hannaneh Hajishirzi; Yulia Tsvetkov; Pradeep Dasigi; Teng Xiao
2026-07-14
2026
arXiv
arXiv
cs.AI
arxiv-api
2607.12227
2026-09-04T05:23:38
ale-0924
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Compaction as Epistemic Failure: How Agentic LLM Tools Fabricate Confirmed Results from Killed Processes
https://arxiv.org/abs/2607.13071
external
arxiv.org
Documents a Claude Code failure in which partial output from a process killed with exit 143 becomes a confirmed claim after context compaction, without re-verification, showing why receipts and process status must survive summarization.
Documents a Claude Code failure in which partial output from a process killed with exit 143 becomes a confirmed claim after context compaction, without re-verification, showing why receipts and process status must survive summarization.
Documents a Claude Code failure in which partial output from a process killed with exit 143 becomes a confirmed claim after context compaction, without re-verification, showing why receipts and process status must survive summarization.
Verification is promoted from a final check to a loop-control signal. Documents a Claude Code failure in which partial output from a process killed with exit 143 becomes a confirmed claim after context compaction, without re-verification, showing why receipts and process status must survive summarization.
Use Compaction as Epistemic Failure: How Agentic LLM Tools Fabricate Confirmed Results from Killed Processes to bound risk before recurring or unattended execution.
Research source arXiv:2607.13071; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,679
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1679
2026-07-17
Govern
govern
Bound permissions, cost, failure, and escalation.
workspace;context;verification;state;exit
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.13071
[2607.13071] Compaction as Epistemic Failure: How Agentic LLM Tools Fabricate Confirmed Results from Killed Processes
Agentic LLM coding tools compress long session histories into compaction summaries that subsequent sessions inherit as ground truth. This paper documents a failure mode in Claude Code where partial standard output from timed-out commands (exit code 143) is recorded in compaction summaries as confirmed results, propagat...
Hiroki Tamba
2026-07-11
2026
arXiv
arXiv
8 pages, companion to arXiv:2606.26185
cs.SE
arxiv-api
2607.13071
2026-09-04T05:23:38
ale-0925
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Do Agent Optimizers Compound? A Continual-Learning Evaluation on Terminal-Bench 2.0
https://arxiv.org/abs/2607.14004
external
arxiv.org
Compares GEPA, Meta Harness, and RELAI-VCL under matched continual-learning budgets; only the regression-controlled method keeps improving, reaching 76.4% lifelong performance versus 66.0%, 64.6%, and 58.7% for the reported alternatives.
Compares GEPA, Meta Harness, and RELAI-VCL under matched continual-learning budgets; only the regression-controlled method keeps improving, reaching 76.4% lifelong performance versus 66.0%, 64.6%, and 58.7% for the reported alternatives.
Compares GEPA, Meta Harness, and RELAI-VCL under matched continual-learning budgets; only the regression-controlled method keeps improving, reaching 76.4% lifelong performance versus 66.0%, 64.6%, and 58.7% for the reported alternatives.
Evaluation data is used as the feedback signal for improving loop behavior. Compares GEPA, Meta Harness, and RELAI-VCL under matched continual-learning budgets; only the regression-controlled method keeps improving, reaching 76.4% lifelong performance versus 66.0%, 64.6%, and 58.7% for the reported alternatives.
Use Do Agent Optimizers Compound? A Continual-Learning Evaluation on Terminal-Bench 2.0 to bound risk before recurring or unattended execution.
Research source arXiv:2607.14004; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,680
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1680
2026-07-17
Govern
govern
Bound permissions, cost, failure, and escalation.
verification;budget
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.14004
[2607.14004] Do Agent Optimizers Compound? A Continual-Learning Evaluation on Terminal-Bench 2.0
Most reported gains from agent-optimization methods are one-shot: an agent is optimized against a fixed benchmark and the resulting improvement is reported as if it were a stable property of the method. This does not test the setting that matters for deployed agents, where optimization is applied recursively as new fai...
Wenxiao Wang; Priyatham Kattakinda; Soheil Feizi
2026-07-15
2026
arXiv
arXiv
Technical Report by RELAI (relai.ai)
cs.AI
arxiv-api
2607.14004
2026-09-04T05:23:38
ale-0926
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Does Multi-Agent Debate Improve AI Feedback on Research Papers?
https://arxiv.org/abs/2607.14713
external
arxiv.org
In a preregistered masked study with authors of 44 meta-analyses, participants prefer single-pass feedback to two multi-agent debate systems, one using about 30x more tokens; AI judges reverse the human preference, warning against self-evaluation alone.
In a preregistered masked study with authors of 44 meta-analyses, participants prefer single-pass feedback to two multi-agent debate systems, one using about 30x more tokens; AI judges reverse the human preference, warning against self-evaluation alone.
In a preregistered masked study with authors of 44 meta-analyses, participants prefer single-pass feedback to two multi-agent debate systems, one using about 30x more tokens; AI judges reverse the human preference, warning against self-evaluation alone.
Evaluation data is used as the feedback signal for improving loop behavior. In a preregistered masked study with authors of 44 meta-analyses, participants prefer single-pass feedback to two multi-agent debate systems, one using about 30x more tokens; AI judges reverse the human preference, warning against self-evaluati...
Use Does Multi-Agent Debate Improve AI Feedback on Research Papers? to bound risk before recurring or unattended execution.
Research source arXiv:2607.14713; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,681
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1681
2026-07-17
Govern
govern
Bound permissions, cost, failure, and escalation.
delegation;verification;budget;escalation
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.14713
[2607.14713] Does Multi-Agent Debate Improve AI Feedback on Research Papers?
Probably not, at least for meta-analyses in economics. In a pre-registered, identity-masked, within-paper experiment, the authors of 44 meta-analyses ranked three AI reports on their own paper by usefulness for improving it: a single pass by a frontier model against two multi-agent debate tools we built and expected to...
Tomas Havranek; Zuzana Irsova
2026-07-16
2026
arXiv
arXiv
29 pages, 1 figure, 6 tables. Pre-registered on OSF; data, code, judge prompts, and blinded reports in the replication package on Zenodo. Project page: https://meta-analysis.cz/debate
econ.GN
arxiv-api
2607.14713
2026-09-04T05:23:38
ale-0927
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Binding Drift in Multi-Step Tool-Augmented Agents
https://arxiv.org/abs/2607.18316
external
arxiv.org
Isolates 'binding drift', entity bindings that start correct then silently go wrong across sequential tool-calling steps, from ordinary error propagation, using 200 workflows and 580 entity-binding-scored steps across four enterprise domains and eight model backends; a naive entity-lock persistence mechanism amplifies ...
Isolates 'binding drift', entity bindings that start correct then silently go wrong across sequential tool-calling steps, from ordinary error propagation, using 200 workflows and 580 entity-binding-scored steps across four enterprise domains and eight model backends; a naive entity-lock persistence mechanism amplifies ...
Isolates 'binding drift', entity bindings that start correct then silently go wrong across sequential tool-calling steps, from ordinary error propagation, using 200 workflows and 580 entity-binding-scored steps across four enterprise domains and eight model backends; a naive entity-lock persistence mechanism amplifies ...
Verification is promoted from a final check to a loop-control signal. Isolates 'binding drift', entity bindings that start correct then silently go wrong across sequential tool-calling steps, from ordinary error propagation, using 200 workflows and 580 entity-binding-scored steps across four enterprise domains and eigh...
Use Binding Drift in Multi-Step Tool-Augmented Agents to bound risk before recurring or unattended execution.
Research source arXiv:2607.18316; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,682
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1682
2026-07-22
Govern
govern
Bound permissions, cost, failure, and escalation.
workspace;verification;state
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.18316
[2607.18316] Binding Drift in Multi-Step Tool-Augmented Agents
Tool-augmented language-model agents execute multi-step workflows over external systems, resolving an entity once and then acting on it across subsequent steps. Prior work shows that in single-step actions, agents select the correct tool but bind it to the wrong entity 24-26% of the time. We study what happens to entit...
Rahul Suresh Babu; Shashank Indukuri
2026-07-17
2026
arXiv
arXiv
14 pages, 5 tables, 1 figure. Equal contribution by both authors. Code and data: https://github.com/shashank-indukuri/binding-drift
cs.SE
arxiv-api
2607.18316
2026-09-04T05:23:38
ale-0928
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
How Agent Skills Fail under Long Contexts: A White-Box Study in Code Auditing
https://arxiv.org/abs/2607.17937
external
arxiv.org
White-box study of skill-following degradation over long tool-using trajectories in a code-audit workflow: pass rates fall from 8/10 in clean context to 3/10 at ~299K characters even though requirement coverage stays above 92%, a detailed external checklist restores 10/10 versus 5/10 for generic self-check, and the pap...
White-box study of skill-following degradation over long tool-using trajectories in a code-audit workflow: pass rates fall from 8/10 in clean context to 3/10 at ~299K characters even though requirement coverage stays above 92%, a detailed external checklist restores 10/10 versus 5/10 for generic self-check, and the pap...
White-box study of skill-following degradation over long tool-using trajectories in a code-audit workflow: pass rates fall from 8/10 in clean context to 3/10 at ~299K characters even though requirement coverage stays above 92%, a detailed external checklist restores 10/10 versus 5/10 for generic self-check, and the pap...
Context is managed as durable loop state rather than a single prompt payload. White-box study of skill-following degradation over long tool-using trajectories in a code-audit workflow: pass rates fall from 8/10 in clean context to 3/10 at ~299K characters even though requirement coverage stays above 92%, a detailed ext...
Use How Agent Skills Fail under Long Contexts: A White-Box Study in Code Auditing to bound risk before recurring or unattended execution.
Research source arXiv:2607.17937; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,683
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1683
2026-07-22
Govern
govern
Bound permissions, cost, failure, and escalation.
workspace;context
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.17937
[2607.17937] When and How Context Rot Appears in Coding Agents: A White-Box Study of Agent Skills in Code Auditing
Agent Skills package procedural instructions and checks for use by general-purpose agents, but loading a skill does not guarantee that every requirement remains active throughout a long tool-using trajectory. We study this problem in a production-derived, white-box code-audit workflow. Holding the task and 24 artifact ...
Yue Xue
2026-07-20
2026
arXiv
arXiv
cs.SE
arxiv-api
2607.17937
2026-09-04T05:23:38
ale-0929
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Phantom Guardrails: When Self-Improving Agent Harnesses Fix Failures That Never Happened
https://arxiv.org/abs/2607.13083
external
arxiv.org
Names a failure mode of self-improving harness loops: the proposer LLM fabricates guardrails for failure classes that provably never occurred (15 of 60 runs when input merely resembles a familiar rule), and once inside an add-only accept loop the phantom guardrail persists; ships a deterministic micro-lab with byte-exa...
Names a failure mode of self-improving harness loops: the proposer LLM fabricates guardrails for failure classes that provably never occurred (15 of 60 runs when input merely resembles a familiar rule), and once inside an add-only accept loop the phantom guardrail persists; ships a deterministic micro-lab with byte-exa...
Names a failure mode of self-improving harness loops: the proposer LLM fabricates guardrails for failure classes that provably never occurred (15 of 60 runs when input merely resembles a familiar rule), and once inside an add-only accept loop the phantom guardrail persists; ships a deterministic micro-lab with byte-exa...
State persistence is explicit enough for repeated runs and handoff. Names a failure mode of self-improving harness loops: the proposer LLM fabricates guardrails for failure classes that provably never occurred (15 of 60 runs when input merely resembles a familiar rule), and once inside an add-only accept loop the phant...
Use Phantom Guardrails: When Self-Improving Agent Harnesses Fix Failures That Never Happened to bound risk before recurring or unattended execution.
Research source arXiv:2607.13083; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,684
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1684
2026-07-22
Govern
govern
Bound permissions, cost, failure, and escalation.
state
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.13083
[2607.13083] Phantom Guardrails: When Self-Improving Agent Harnesses Fix Failures That Never Happened
Self-improving AI agents are designed to learn from their mistakes. We show they can also hallucinate mistakes that never happened. We study this failure mode in automated harness optimization, where an LLM-based proposer edits an agent's scaffold, including prompts, parsers, filters, validators and guardrails, to elim...
Su Wang; Pin Qian; Yifan Lin; Jingzhou Xu; Yihang Chen; Xiaochong Jiang; Lifei Liu; Haoran Yu
2026-07-13
2026
arXiv
arXiv
cs.CR
arxiv-api
2607.13083
2026-09-04T05:23:38
ale-0930
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Skills That Don't Exist: A Large-Scale Study of Hallucinated Skill Recommendation
https://arxiv.org/abs/2607.12340
external
arxiv.org
Measures a 36% average rate of agents recommending non-existent skills across 15,000 prompts, and shows the same fake names recur consistently, enabling slopsquatting-style supply-chain attacks where adversaries pre-register malicious skills under the hallucinated names agents will predictably ask for.
Measures a 36% average rate of agents recommending non-existent skills across 15,000 prompts, and shows the same fake names recur consistently, enabling slopsquatting-style supply-chain attacks where adversaries pre-register malicious skills under the hallucinated names agents will predictably ask for.
Measures a 36% average rate of agents recommending non-existent skills across 15,000 prompts, and shows the same fake names recur consistently, enabling slopsquatting-style supply-chain attacks where adversaries pre-register malicious skills under the hallucinated names agents will predictably ask for.
Keeps adoption grounded in known failure modes, economics, and operational limits. Measures a 36% average rate of agents recommending non-existent skills across 15,000 prompts, and shows the same fake names recur consistently, enabling slopsquatting-style supply-chain attacks where adversaries pre-register malicious sk...
Use Skills That Don't Exist: A Large-Scale Study of Hallucinated Skill Recommendation to bound risk before recurring or unattended execution.
Research source arXiv:2607.12340; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,685
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1685
2026-07-22
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.12340
[2607.12340] Skills That Don't Exist: A Large-Scale Study of Hallucinated Skill Recommendation in LLM Agents
LLM agents acquire new capabilities by downloading skills from open registries. Instead of browsing these catalogs manually, developers typically ask the agent to recommend and install a skill. This convenience hides a risk: agents frequently invent names for skills that exist in no registry. We term this flaw skill na...
Weifeng Yuan; Wenbo Guo; Feng Dong; Haoyu Wang; Yang Liu
2026-07-14
2026
arXiv
arXiv
cs.SE
arxiv-api
2607.12340
2026-09-04T05:23:38
ale-0931
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Token Reduction Is Not Cost Reduction
https://arxiv.org/abs/2607.12161
external
arxiv.org
Analysis of 2,848 real Claude Code runs showing prompt-cache traffic accounts for ~87% of billed cost, so local token/context compression does not reliably lower the bill, argues loop cost engineering should optimize success-adjusted billed cost, not token counts. Directly actionable for anyone running high-volume agen...
Analysis of 2,848 real Claude Code runs showing prompt-cache traffic accounts for ~87% of billed cost, so local token/context compression does not reliably lower the bill, argues loop cost engineering should optimize success-adjusted billed cost, not token counts. Directly actionable for anyone running high-volume agen...
Analysis of 2,848 real Claude Code runs showing prompt-cache traffic accounts for ~87% of billed cost, so local token/context compression does not reliably lower the bill, argues loop cost engineering should optimize success-adjusted billed cost, not token counts. Directly actionable for anyone running high-volume agen...
Context is managed as durable loop state rather than a single prompt payload. Analysis of 2,848 real Claude Code runs showing prompt-cache traffic accounts for ~87% of billed cost, so local token/context compression does not reliably lower the bill, argues loop cost engineering should optimize success-adjusted billed c...
Use Token Reduction Is Not Cost Reduction to bound risk before recurring or unattended execution.
Research source arXiv:2607.12161; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,686
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1686
2026-07-22
Govern
govern
Bound permissions, cost, failure, and escalation.
context;budget
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.12161
[2607.12161] Token Reduction Is Not Cost Reduction
Token-reduction tools for coding agents are often evaluated by the number of tokens they remove, but token count alone does not determine end-to-end inference cost. We evaluate three token-reduction approaches against an unmodified Claude Code baseline across controlled coding tasks, measuring provider-billed cost, tas...
Sarel Weinberger; Amir Hozez
2026-07-13
2026
arXiv
arXiv
cs.CL
arxiv-api
2607.12161
2026-09-04T05:23:38
ale-0932
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Blog
📝
What xAI's Grok Build CLI Actually Sends: A Wire-Level Analysis
https://gist.github.com/cereblab/dc9a40bc26120f4540e4e09b75ffb547
external
gist.github.com
Wire-level analysis of the telemetry and payloads Grok Build transmits, a reminder that agent harnesses carry their own data-flow surface worth auditing before unattended use.
Wire-level analysis of the telemetry and payloads Grok Build transmits, a reminder that agent harnesses carry their own data-flow surface worth auditing before unattended use.
Wire-level analysis of the telemetry and payloads Grok Build transmits, a reminder that agent harnesses carry their own data-flow surface worth auditing before unattended use.
Keeps adoption grounded in known failure modes, economics, and operational limits. Wire-level analysis of the telemetry and payloads Grok Build transmits, a reminder that agent harnesses carry their own data-flow surface worth auditing before unattended use.
Use What xAI's Grok Build CLI Actually Sends: A Wire-Level Analysis to bound risk before recurring or unattended execution.
Contextual source from gist.github.com; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,687
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1687
2026-07-22
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
operator;security
cross-layer
enabling
practitioner-analysis
B
ok
https://gist.github.com/cereblab/dc9a40bc26120f4540e4e09b75ffb547
What xAI Grok Build CLI actually sends to xAI - a wire-level analysis (grok 0.2.93) · GitHub
What xAI Grok Build CLI actually sends to xAI - a wire-level analysis (grok 0.2.93) - grok-build-cli-wire-analysis.md
Gist
html-meta
2026-09-04T05:23:38
ale-0933
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Critique
⚠️
The Tower Keeps Rising
https://lucumr.pocoo.org/2026/7/13/the-tower-keeps-rising/
external
lucumr.pocoo.org
Armin Ronacher's follow-up to The Coming Loop, on abstraction layers accumulating faster than understanding as agent tooling stacks up, and what that does to a codebase's long-term comprehensibility.
Armin Ronacher's follow-up to The Coming Loop, on abstraction layers accumulating faster than understanding as agent tooling stacks up, and what that does to a codebase's long-term comprehensibility.
Armin Ronacher's follow-up to The Coming Loop, on abstraction layers accumulating faster than understanding as agent tooling stacks up, and what that does to a codebase's long-term comprehensibility.
Keeps adoption grounded in known failure modes, economics, and operational limits. Armin Ronacher's follow-up to The Coming Loop, on abstraction layers accumulating faster than understanding as agent tooling stacks up, and what that does to a codebase's long-term comprehensibility.
Use The Tower Keeps Rising to bound risk before recurring or unattended execution.
Contextual source from lucumr.pocoo.org; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,688
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1688
2026-07-22
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
operator;security
cross-layer
enabling
risk-analysis
B
ok
https://lucumr.pocoo.org/2026/7/13/the-tower-keeps-rising/
The Tower Keeps Rising | Armin Ronacher's Thoughts and Writings
Vibecoding and the possible collapse of a shared language.
2026-07-13
2026
Armin Ronacher's Thoughts and Writings
html-meta
2026-09-04T05:23:38
ale-0934
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
The Dark Room in the Reward Channel: Dense Prediction Rewards Collapse GRPO-Trained LLM Agents -- and What Actually Works
https://arxiv.org/abs/2607.21273
external
arxiv.org
Shows dense next-observation prediction rewards under GRPO drive long-horizon LLM agents into a degenerate "dark room" absorbing state (prediction accuracy 1.0, task success 0), with a single-factor ablation localizing the cause to std normalization and auxiliary-loss channels as a working fix. A crisp negative result ...
Shows dense next-observation prediction rewards under GRPO drive long-horizon LLM agents into a degenerate "dark room" absorbing state (prediction accuracy 1.0, task success 0), with a single-factor ablation localizing the cause to std normalization and auxiliary-loss channels as a working fix. A crisp negative result ...
Shows dense next-observation prediction rewards under GRPO drive long-horizon LLM agents into a degenerate "dark room" absorbing state (prediction accuracy 1.0, task success 0), with a single-factor ablation localizing the cause to std normalization and auxiliary-loss channels as a working fix. A crisp negative result ...
The work targets tasks that exceed a single context window or prompt session. Shows dense next-observation prediction rewards under GRPO drive long-horizon LLM agents into a degenerate "dark room" absorbing state (prediction accuracy 1.0, task success 0), with a single-factor ablation localizing the cause to std normal...
Use The Dark Room in the Reward Channel: Dense Prediction Rewards Collapse GRPO-Trained LLM Agents -- and What Actually Works to bound risk before recurring or unattended execution.
Research source arXiv:2607.21273; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,689
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1689
2026-07-24
Govern
govern
Bound permissions, cost, failure, and escalation.
state
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.21273
[2607.21273] The Dark Room in the Reward Channel: Dense Prediction Rewards Collapse GRPO-Trained LLM Agents -- and The Channel, Not the Content, Decides What Works
Dense per-step supervision is the standard remedy for sparse-reward long-horizon LLM agents: reward the policy for predicting its next observation, which looks provably safe under potential-based shaping. Published prediction-reward and auxiliary-loss variants report both successes and instabilities; we supply the cont...
Yu Wang
2026-07-23
2026
arXiv
arXiv
10.5281/zenodo.21505228
cs.LG
arxiv-api
2607.21273
2026-09-04T05:23:38
ale-0935
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Blog
📝
Why Software Factories Fail (or: Harness Engineering Is Not Enough)
https://github.com/humanlayer/advanced-context-engineering-for-coding-agents/blob/main/wsff.md
external
github.com
Dex Horthy's essay from his AI Engineer World's Fair 2026 keynote arguing that lights-off software factories fail because RL training rewards passing tests with no penalty for eroding codebase maintainability, so long-horizon quality feedback cannot be trained on, proposing front-loaded human judgment gates (requiremen...
Dex Horthy's essay from his AI Engineer World's Fair 2026 keynote arguing that lights-off software factories fail because RL training rewards passing tests with no penalty for eroding codebase maintainability, so long-horizon quality feedback cannot be trained on, proposing front-loaded human judgment gates (requiremen...
Dex Horthy's essay from his AI Engineer World's Fair 2026 keynote arguing that lights-off software factories fail because RL training rewards passing tests with no penalty for eroding codebase maintainability, so long-horizon quality feedback cannot be trained on, proposing front-loaded human judgment gates (requiremen...
The work targets tasks that exceed a single context window or prompt session. Dex Horthy's essay from his AI Engineer World's Fair 2026 keynote arguing that lights-off software factories fail because RL training rewards passing tests with no penalty for eroding codebase maintainability, so long-horizon quality feedback...
Use Why Software Factories Fail (or: Harness Engineering Is Not Enough) to bound risk before recurring or unattended execution.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,690
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1690
2026-07-24
Govern
govern
Bound permissions, cost, failure, and escalation.
verification;escalation
operator;security
cross-layer
enabling
practitioner-analysis
B
ok
https://github.com/humanlayer/advanced-context-engineering-for-coding-agents/blob/main/wsff.md
advanced-context-engineering-for-coding-agents/wsff.md at main · humanlayer/advanced-context-engineering-for-coding-agents · GitHub
Contribute to humanlayer/advanced-context-engineering-for-coding-agents development by creating an account on GitHub.
humanlayer/advanced-context-engineering-for-coding-agents
GitHub
html-meta
humanlayer/advanced-context-engineering-for-coding-agents
2026-09-04T05:23:38
ale-0936
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
The Boundaries of Automation: A Theory of Persistent Human Participation
https://arxiv.org/abs/2607.21547
external
arxiv.org
Theory paper from Fourati, Schütze, Hüllermeier, and Gurevych challenging the assumption that humans stay in the loop only until AI capability catches up: it identifies three grounds for persistent human participation, technical complementarity, normative/developmental value, and objectives that emerge through the inte...
Theory paper from Fourati, Schütze, Hüllermeier, and Gurevych challenging the assumption that humans stay in the loop only until AI capability catches up: it identifies three grounds for persistent human participation, technical complementarity, normative/developmental value, and objectives that emerge through the inte...
Theory paper from Fourati, Schütze, Hüllermeier, and Gurevych challenging the assumption that humans stay in the loop only until AI capability catches up: it identifies three grounds for persistent human participation, technical complementarity, normative/developmental value, and objectives that emerge through the inte...
State persistence is explicit enough for repeated runs and handoff. Theory paper from Fourati, Schütze, Hüllermeier, and Gurevych challenging the assumption that humans stay in the loop only until AI capability catches up: it identifies three grounds for persistent human participation, technical complementarity, normat...
Use The Boundaries of Automation: A Theory of Persistent Human Participation to bound risk before recurring or unattended execution.
Research source arXiv:2607.21547; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,691
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1691
2026-07-25
Govern
govern
Bound permissions, cost, failure, and escalation.
objective;state;escalation
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.21547
[2607.21547] The Boundaries of Automation: A Theory of Persistent Human Participation
The rapid progress of AI has intensified the long-standing pursuit of automation: replacing human participation with algorithms wherever possible. Implicit in this pursuit is the assumption that humans remain in the loop only because current AI systems are not yet sufficiently capable. This paper challenges that assump...
Fares Fourati; Hinrich Schütze; Eyke Hüllermeier; Iryna Gurevych
2026-07-23
2026
arXiv
arXiv
cs.AI
arxiv-api
2607.21547
2026-09-04T05:23:38
ale-0937
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Tool
🧰
Reward Hacking in the Wild
https://rewardhacking.org
external
rewardhacking.org
Searchable corpus of 3,607 user-reported AI-agent misbehavior incidents collected from GitHub, Hacker News, LessWrong, and X, normalized and classified across fourteen misbehavior types with severity ratings.
Searchable corpus of 3,607 user-reported AI-agent misbehavior incidents collected from GitHub, Hacker News, LessWrong, and X, normalized and classified across fourteen misbehavior types with severity ratings.
Searchable corpus of 3,607 user-reported AI-agent misbehavior incidents collected from GitHub, Hacker News, LessWrong, and X, normalized and classified across fourteen misbehavior types with severity ratings.
Keeps adoption grounded in known failure modes, economics, and operational limits. Searchable corpus of 3,607 user-reported AI-agent misbehavior incidents collected from GitHub, Hacker News, LessWrong, and X, normalized and classified across fourteen misbehavior types with severity ratings.
Use Reward Hacking in the Wild to bound risk before recurring or unattended execution.
Working implementation, framework, runtime, or repository; signal comes from usable code and ecosystem adoption.
high
README.md
1,692
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1692
2026-07-25
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
builder;operator;security
cross-layer
enabling
implementation
A
ok
https://rewardhacking.org
Your AIs don't do what you want. This is really bad negligible: 1,468 (40.7%) minor: 1,373 (38.1%) significant: 618 (17.1%) severe: 121 (3.4%) unrated: 27 (0.7%)
Thousands of user-reported incidents of AI agents misbehaving, collected from public posts. Reports, not verified events.
Reward Hacking in the Wild
html-meta
2026-09-04T05:23:38
ale-0938
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
What AI Red-Team Evaluations Can and Cannot Prove
https://arxiv.org/abs/2607.21735
external
arxiv.org
Bandana Kaur gives a formal treatment of the epistemic limits of red-teaming, the practice now serving as the industry's primary safety gate for agent deployment.
Bandana Kaur gives a formal treatment of the epistemic limits of red-teaming, the practice now serving as the industry's primary safety gate for agent deployment.
Bandana Kaur gives a formal treatment of the epistemic limits of red-teaming, the practice now serving as the industry's primary safety gate for agent deployment.
Keeps adoption grounded in known failure modes, economics, and operational limits. Bandana Kaur gives a formal treatment of the epistemic limits of red-teaming, the practice now serving as the industry's primary safety gate for agent deployment.
Use What AI Red-Team Evaluations Can and Cannot Prove to bound risk before recurring or unattended execution.
Research source arXiv:2607.21735; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,693
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1693
2026-07-28
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.21735
[2607.21735] What AI Red-Team Evaluations Can and Cannot Prove
Red-team evaluations of AI models support some claims and not others, and the boundary between the two is calculable rather than merely a matter of judgment. We define the evidential ceiling of an evaluation as the largest factor by which one result can move belief under a fixed testing budget, derive it in closed form...
Bandana Kaur
2026-07-23
2026
arXiv
arXiv
21 pages, 4 figures, 5 tables. Code and data links provided in the manuscript. v2: corrected Figure 1(b); corrected required sample sizes in Table 4 and in Sections 4.2, 4.6 and 5.2, which had been rounded rather than taken to the ceiling; corrected the sample-size expression stated in Methods; minor corrections to Tab...
cs.AI
arxiv-api
2607.21735
2026-09-04T05:23:38
ale-0939
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
The Best Programming Language for Tokenmaxxing: An Investigation of Coding Agent Behavior Across Programming Languages
https://arxiv.org/abs/2607.22807
external
arxiv.org
Shows coding-agent token cost varies starkly and consistently by programming language across five recent models on difficulty-controlled Python, Java, Rust, and OCaml problems.
Shows coding-agent token cost varies starkly and consistently by programming language across five recent models on difficulty-controlled Python, Java, Rust, and OCaml problems.
Shows coding-agent token cost varies starkly and consistently by programming language across five recent models on difficulty-controlled Python, Java, Rust, and OCaml problems.
Keeps adoption grounded in known failure modes, economics, and operational limits. Shows coding-agent token cost varies starkly and consistently by programming language across five recent models on difficulty-controlled Python, Java, Rust, and OCaml problems.
Use The Best Programming Language for Tokenmaxxing: An Investigation of Coding Agent Behavior Across Programming Languages to bound risk before recurring or unattended execution.
Research source arXiv:2607.22807; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,694
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1694
2026-07-28
Govern
govern
Bound permissions, cost, failure, and escalation.
budget
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.22807
[2607.22807] The Best Programming Language for Tokenmaxxing: An Investigation of Coding Agent Behavior Across Programming Languages
Although coding agents are now very effective in a variety of programming languages, this paper first shows that the cost (in tokens) can very significantly by programming language. We evaluate five recent models on programming problems in Python, Java, Rust, and OCaml. We carefully control for problem difficulty, and ...
Zixuan Wu; Carolyn Jane Anderson; Arjun Guha
2026-07-24
2026
arXiv
arXiv
cs.SE
arxiv-api
2607.22807
2026-09-04T05:23:38
ale-0940
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Where Is the Cost of Third-Party API Routers in Agentic Software Development?
https://arxiv.org/abs/2607.23624
external
arxiv.org
Third-party LLM routers sit on the trusted path between a coding agent and its provider, able to inspect and modify every request and response, while nothing verifies that provider output matches the repository-level actions the agent ultimately executes -- so client-side permission mechanisms can silently stop working...
Third-party LLM routers sit on the trusted path between a coding agent and its provider, able to inspect and modify every request and response, while nothing verifies that provider output matches the repository-level actions the agent ultimately executes -- so client-side permission mechanisms can silently stop working...
Third-party LLM routers sit on the trusted path between a coding agent and its provider, able to inspect and modify every request and response, while nothing verifies that provider output matches the repository-level actions the agent ultimately executes -- so client-side permission mechanisms can silently stop working...
Keeps adoption grounded in known failure modes, economics, and operational limits. Third-party LLM routers sit on the trusted path between a coding agent and its provider, able to inspect and modify every request and response, while nothing verifies that provider output matches the repository-level actions the agent ul...
Use Where Is the Cost of Third-Party API Routers in Agentic Software Development? to bound risk before recurring or unattended execution.
Research source arXiv:2607.23624; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,695
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1695
2026-07-28
Govern
govern
Bound permissions, cost, failure, and escalation.
workspace;verification;budget;exit
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.23624
[2607.23624] Where Is the Cost of Third-Party API Routers in Agentic Software Development?
Third-party API routers have become a common layer that unifies access across increasingly diverse LLM providers. In coding-agent workflows, high-autonomy operation is widely adopted because it reduces interaction overhead. As a result, a third-party API router, which sits between the agent and the upstream provider, i...
Donghao Fu; Jingxin Li; Xue Jiang; Yihong Dong
2026-07-26
2026
arXiv
arXiv
cs.SE
arxiv-api
2607.23624
2026-09-04T05:23:38
ale-0941
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Efficiency Matters in Autonomous Research
https://arxiv.org/abs/2607.24647
external
arxiv.org
Position paper arguing autonomous research systems are judged almost entirely on final outcome quality while search efficiency -- reaching that outcome on a small budget -- is an equally important and ignored dimension, and one that dominates as AR moves from cheap-verification domains like math and code into settings ...
Position paper arguing autonomous research systems are judged almost entirely on final outcome quality while search efficiency -- reaching that outcome on a small budget -- is an equally important and ignored dimension, and one that dominates as AR moves from cheap-verification domains like math and code into settings ...
Position paper arguing autonomous research systems are judged almost entirely on final outcome quality while search efficiency -- reaching that outcome on a small budget -- is an equally important and ignored dimension, and one that dominates as AR moves from cheap-verification domains like math and code into settings ...
Verification is promoted from a final check to a loop-control signal. Position paper arguing autonomous research systems are judged almost entirely on final outcome quality while search efficiency -- reaching that outcome on a small budget -- is an equally important and ignored dimension, and one that dominates as AR m...
Use Efficiency Matters in Autonomous Research to bound risk before recurring or unattended execution.
Research source arXiv:2607.24647; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,696
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1696
2026-07-28
Govern
govern
Bound permissions, cost, failure, and escalation.
verification;budget
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.24647
[2607.24647] Efficiency Matters in Autonomous Research
AI-driven autonomous research (AR) systems are becoming increasingly effective across a broad range of tasks. Their performance, however, is still evaluated primarily by the quality of the final outcome. In this paper, we argue that the efficiency of the solution-search process is an equally important but often overloo...
Haiqian Yang; Yuan Cao
2026-07-27
2026
arXiv
arXiv
cs.AI
arxiv-api
2607.24647
2026-09-04T05:23:38
ale-0942
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Reliability-Contagion Feasibility in LLM Multi-Agent Networks
https://arxiv.org/abs/2607.21912
external
arxiv.org
Treats error propagation in multi-agent systems as an epidemic on the communication graph, with susceptible/exposed/infectious/corrected states and a derived early-invasion condition for heterogeneous topologies, then couples it to an analytic majority-vote benchmark where a clean-task reliability target imposes a mini...
Treats error propagation in multi-agent systems as an epidemic on the communication graph, with susceptible/exposed/infectious/corrected states and a derived early-invasion condition for heterogeneous topologies, then couples it to an analytic majority-vote benchmark where a clean-task reliability target imposes a mini...
Treats error propagation in multi-agent systems as an epidemic on the communication graph, with susceptible/exposed/infectious/corrected states and a derived early-invasion condition for heterogeneous topologies, then couples it to an analytic majority-vote benchmark where a clean-task reliability target imposes a mini...
Control flow is represented as an inspectable graph rather than an opaque prompt loop. Treats error propagation in multi-agent systems as an epidemic on the communication graph, with susceptible/exposed/infectious/corrected states and a derived early-invasion condition for heterogeneous topologies, then couples it to a...
Use Reliability-Contagion Feasibility in LLM Multi-Agent Networks to bound risk before recurring or unattended execution.
Research source arXiv:2607.21912; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,697
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1697
2026-07-28
Govern
govern
Bound permissions, cost, failure, and escalation.
delegation;verification
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.21912
[2607.21912] Reliability-Contagion Feasibility in LLM Multi-Agent Networks
Communication allows large language model agents to pool evidence, but it also creates paths along which an erroneous claim can spread. We formulate a correction-aware network model that tracks susceptible, exposed, infectious, and corrected agents and derive its early-invasion condition for heterogeneous communication...
Ruiwu Niu; Xincheng Shu; Ying Zhao
2026-07-24
2026
arXiv
arXiv
cs.MA
arxiv-api
2607.21912
2026-09-04T05:23:38
ale-0943
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Draining the Energy Commons: Self-Defeating Over-Appropriation as a Coordination Failure in Agentic LLM Collectives
https://arxiv.org/abs/2607.22188
external
arxiv.org
When LLM agents share a persistent resource, one agent's decision changes the conditions later agents face. Four same-family GPT, Gemini, or Grok agents act as electricity prosumers instructed to maximize operational continuity, with aggregate demand and protocol held fixed while the regeneration rate varies.
When LLM agents share a persistent resource, one agent's decision changes the conditions later agents face. Four same-family GPT, Gemini, or Grok agents act as electricity prosumers instructed to maximize operational continuity, with aggregate demand and protocol held fixed while the regeneration rate varies.
When LLM agents share a persistent resource, one agent's decision changes the conditions later agents face. Four same-family GPT, Gemini, or Grok agents act as electricity prosumers instructed to maximize operational continuity, with aggregate demand and protocol held fixed while the regeneration rate varies.
State persistence is explicit enough for repeated runs and handoff. When LLM agents share a persistent resource, one agent's decision changes the conditions later agents face. Four same-family GPT, Gemini, or Grok agents act as electricity prosumers instructed to maximize operational continuity, with aggregate demand a...
Use Draining the Energy Commons: Self-Defeating Over-Appropriation as a Coordination Failure in Agentic LLM Collectives to bound risk before recurring or unattended execution.
Research source arXiv:2607.22188; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,698
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1698
2026-07-28
Govern
govern
Bound permissions, cost, failure, and escalation.
state
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.22188
[2607.22188] Draining the Energy Commons: Self-Defeating Over-Appropriation as a Coordination Failure in Agentic LLM Collectives
LLMs are increasingly deployed as agents that plan, use tools, and act over time. When they share persistent resources, such as compute pools or energy reserves, decisions by one agent affect the conditions faced by later agents. We study this coordination failure in a renewable energy commons. Four same-family GPT, Ge...
Marcantonio Bracale Syrnikov; Federico Pierucci; Matteo Prandi; Marcello Galisai; Piercosma Bisconti; Francesco Giarrusso; Daniele Nardi
2026-07-24
2026
arXiv
arXiv
cs.MA
arxiv-api
2607.22188
2026-09-04T05:23:38
ale-0944
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
"Go Home Copilot, You're Drunk": Understanding Developer Responses to Agent-Generated Code Review Comments
https://arxiv.org/abs/2607.21997
external
arxiv.org
First large-scale empirical study of what happens after an agent posts a review comment: 54,791 comments from Copilot, Cursor, Codex, Devin, and Claude across 342 Python GitHub repositories, analyzed for resolution rates by agent and comment type, the effect of developer experience, and what makes a comment useful.
First large-scale empirical study of what happens after an agent posts a review comment: 54,791 comments from Copilot, Cursor, Codex, Devin, and Claude across 342 Python GitHub repositories, analyzed for resolution rates by agent and comment type, the effect of developer experience, and what makes a comment useful.
First large-scale empirical study of what happens after an agent posts a review comment: 54,791 comments from Copilot, Cursor, Codex, Devin, and Claude across 342 Python GitHub repositories, analyzed for resolution rates by agent and comment type, the effect of developer experience, and what makes a comment useful.
Keeps adoption grounded in known failure modes, economics, and operational limits. First large-scale empirical study of what happens after an agent posts a review comment: 54,791 comments from Copilot, Cursor, Codex, Devin, and Claude across 342 Python GitHub repositories, analyzed for resolution rates by agent and com...
Use "Go Home Copilot, You're Drunk": Understanding Developer Responses to Agent-Generated Code Review Comments to bound risk before recurring or unattended execution.
Research source arXiv:2607.21997; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,699
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1699
2026-07-28
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.21997
[2607.21997] "Go Home Copilot, You're Drunk": Understanding Developer Responses to Agent-Generated Code Review Comments
Code review is a critical quality assurance practice in software engineering development, and AI coding agents are increasingly generating review comments on pull requests. However, little is known about how developers actually respond to such agent-generated feedback. In this paper, we present the first large-scale em...
Shamse Tasnim Cynthia; Ratnadira Widyasari; Banani Roy; Ting Zhang; David Lo
2026-07-24
2026
arXiv
arXiv
cs.SE
arxiv-api
2607.21997
2026-09-04T05:23:38
ale-0945
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Position: Evaluation Scores Are Perishable Knowledge Claims
https://arxiv.org/abs/2607.26191
external
arxiv.org
Argues benchmark scores are epistemic claims with expiry dates and that averaging heterogeneous signals inflates confidence past the weakest component, proposing formality tiers, scope declarations, and expirations, HELM rankings shift materially under weakest-link aggregation. Sharp critique of how loop teams read eva...
Argues benchmark scores are epistemic claims with expiry dates and that averaging heterogeneous signals inflates confidence past the weakest component, proposing formality tiers, scope declarations, and expirations, HELM rankings shift materially under weakest-link aggregation. Sharp critique of how loop teams read eva...
Argues benchmark scores are epistemic claims with expiry dates and that averaging heterogeneous signals inflates confidence past the weakest component, proposing formality tiers, scope declarations, and expirations, HELM rankings shift materially under weakest-link aggregation. Sharp critique of how loop teams read eva...
Evaluation data is used as the feedback signal for improving loop behavior. Argues benchmark scores are epistemic claims with expiry dates and that averaging heterogeneous signals inflates confidence past the weakest component, proposing formality tiers, scope declarations, and expirations, HELM rankings shift material...
Use Position: Evaluation Scores Are Perishable Knowledge Claims to bound risk before recurring or unattended execution.
Research source arXiv:2607.26191; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,700
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1700
2026-07-30
Govern
govern
Bound permissions, cost, failure, and escalation.
verification
researcher;evaluator;operator;security
cross-layer
enabling
research-paper
A
ok
https://doi.org/10.18653/v1/2026.gem-main.80
[2607.26191] Position: Evaluation Scores Are Perishable Knowledge Claims
Evaluation methodologies for language models increasingly combine multiple signals, from automated metrics and LLM-as-judge ratings to human assessments and benchmark suite results. When these signals are aggregated via averaging, evaluation confidence can then substantially exceed the reliability of the weakest signal...
Sankalp Gilda; Shlok Gilda
2026
2026
Proceedings of the Fifth Workshop on Generation, Evaluation and Metrics (GEM) 2026
Association for Computational Linguistics
10.18653/v1/2026.gem-main.80
Published in Proceedings of the Fifth Workshop on Generation, Evaluation and Metrics (GEM) 2026; the linked arXiv record remains available for open access.
cs.AI
Crossref API + DOI record
2607.26191
2026-09-04T05:23:38
ale-0946
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
One Run Is Not an Idea: The Implementation Lottery in Automated Research
https://arxiv.org/abs/2607.26587
external
arxiv.org
Shows automated research loops judge ideas on a single implementation, that implementation variance dwarfs rerun variance, and that winner reversal rates exceed 25%, then offers an Idea Reliability Audit. Directly undermines single-run selection in self-improving research agents.
Shows automated research loops judge ideas on a single implementation, that implementation variance dwarfs rerun variance, and that winner reversal rates exceed 25%, then offers an Idea Reliability Audit. Directly undermines single-run selection in self-improving research agents.
Shows automated research loops judge ideas on a single implementation, that implementation variance dwarfs rerun variance, and that winner reversal rates exceed 25%, then offers an Idea Reliability Audit. Directly undermines single-run selection in self-improving research agents.
Keeps adoption grounded in known failure modes, economics, and operational limits. Shows automated research loops judge ideas on a single implementation, that implementation variance dwarfs rerun variance, and that winner reversal rates exceed 25%, then offers an Idea Reliability Audit. Directly undermines single-run s...
Use One Run Is Not an Idea: The Implementation Lottery in Automated Research to bound risk before recurring or unattended execution.
Research source arXiv:2607.26587; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,701
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1701
2026-07-30
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.26587
[2607.26587] One Run Is Not an Idea: The Implementation Lottery in Automated Research
Automated research systems use experimental scores both to deliver artifacts and to decide which ideas to retain, transfer, and pursue. Yet one run scores one implementation of an idea. Crediting that realization-level score as evidence about the parent mechanism creates the \emph{implementation lottery}, in which an i...
Jingjie Ning; Shanshan Zhong; Xiaochuan Li; Ji Zeng; Chenyan Xiong
2026-07-29
2026
arXiv
arXiv
cs.MA
arxiv-api
2607.26587
2026-09-04T05:23:38
ale-0947
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Can AI agents conduct open-ended AI research? Early evidence from two case studies
https://arxiv.org/abs/2607.27191
external
arxiv.org
Frontier agents ran six days of largely independent engineering on open research questions from unpublished NeurIPS 2026 papers, and the original authors unambiguously rejected both outputs, the gap was research judgment and creativity, not execution. A rigorous shadow-evaluation counterweight to autonomous-research op...
Frontier agents ran six days of largely independent engineering on open research questions from unpublished NeurIPS 2026 papers, and the original authors unambiguously rejected both outputs, the gap was research judgment and creativity, not execution. A rigorous shadow-evaluation counterweight to autonomous-research op...
Frontier agents ran six days of largely independent engineering on open research questions from unpublished NeurIPS 2026 papers, and the original authors unambiguously rejected both outputs, the gap was research judgment and creativity, not execution. A rigorous shadow-evaluation counterweight to autonomous-research op...
Evaluation data is used as the feedback signal for improving loop behavior. Frontier agents ran six days of largely independent engineering on open research questions from unpublished NeurIPS 2026 papers, and the original authors unambiguously rejected both outputs, the gap was research judgment and creativity, not exe...
Use Can AI agents conduct open-ended AI research? Early evidence from two case studies to bound risk before recurring or unattended execution.
Research source arXiv:2607.27191; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,702
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1702
2026-07-30
Govern
govern
Bound permissions, cost, failure, and escalation.
verification
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.27191
[2607.27191] Can AI agents conduct open-ended AI research? Early evidence from two case studies
Forecasts of explosive AI progress hinge on AI agents automating AI research. But evidence on whether agents can carry out open-ended AI research is thin. Current evaluations either test agents on narrow, verifiable tasks, which excludes open-ended research, or submit AI-generated papers to blind peer review, which is ...
Peter Kirgis; Sayash Kapoor; Andrew Schwartz; Stephan Rabanser; David Africa; Konstantinos Voudouris; Viet Nguyen; Toby Pilditch; Magda Dubois; Harry Coppock; Cozmin Ududec; Nitya Nadgir; Matilda Orona; Tilman Bayer; Derrick Chan-Sew; Yue Ling; Abhishek Shetty; Helen Toner; Gillian Hadfield; Seth Lazar; Steve Newman; S...
2026-07-29
2026
arXiv
arXiv
cs.AI
arxiv-api
2607.27191
2026-09-04T05:23:38
ale-0948
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Two Calls Beat Five Agents: Evaluating Multi-Agent Pipelines Against Self-Refinement for Local Language Models
https://arxiv.org/abs/2607.26922
external
arxiv.org
On a 7B local model, two refinement iterations beat a five-agent architecture on math reasoning, with data formatting and task-specific tuning mattering more than topology. Practical argument against reaching for multi-agent structure before exhausting the simple loop.
On a 7B local model, two refinement iterations beat a five-agent architecture on math reasoning, with data formatting and task-specific tuning mattering more than topology. Practical argument against reaching for multi-agent structure before exhausting the simple loop.
On a 7B local model, two refinement iterations beat a five-agent architecture on math reasoning, with data formatting and task-specific tuning mattering more than topology. Practical argument against reaching for multi-agent structure before exhausting the simple loop.
The work separates roles across agents, verifiers, or orchestration layers. On a 7B local model, two refinement iterations beat a five-agent architecture on math reasoning, with data formatting and task-specific tuning mattering more than topology. Practical argument against reaching for multi-agent structure before ex...
Use Two Calls Beat Five Agents: Evaluating Multi-Agent Pipelines Against Self-Refinement for Local Language Models to bound risk before recurring or unattended execution.
Research source arXiv:2607.26922; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,703
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1703
2026-07-30
Govern
govern
Bound permissions, cost, failure, and escalation.
delegation
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.26922
[2607.26922] Two Calls Beat Five Agents: Evaluating Multi-Agent Pipelines Against Self-Refinement for Local Language Models
Multi-agent LLM pipeline systems break down the task among multiple roles for better reasoning, but are benchmarked mainly with large-scale commercial models. In this study, we investigate Parishad, a structured multi-agent system involving five roles, by deploying it on Qwen2.5-7B-Instruct, a local model, on two datas...
Ashish Prajapati; Om Mohite
2026-07-29
2026
arXiv
arXiv
cs.LG
arxiv-api
2607.26922
2026-09-04T05:23:38
ale-0949
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Try Again, Don't Look Back: Blind Resampling Outperforms Self-Repair in Small Code Models
https://arxiv.org/abs/2607.26117
external
arxiv.org
Finds that resampling without showing the model its failed attempt beats self-repair, because conditioning on its own broken code anchors the model into reproducing near-identical flaws. A concrete case where the feedback in the feedback loop actively hurts.
Finds that resampling without showing the model its failed attempt beats self-repair, because conditioning on its own broken code anchors the model into reproducing near-identical flaws. A concrete case where the feedback in the feedback loop actively hurts.
Finds that resampling without showing the model its failed attempt beats self-repair, because conditioning on its own broken code anchors the model into reproducing near-identical flaws. A concrete case where the feedback in the feedback loop actively hurts.
Keeps adoption grounded in known failure modes, economics, and operational limits. Finds that resampling without showing the model its failed attempt beats self-repair, because conditioning on its own broken code anchors the model into reproducing near-identical flaws. A concrete case where the feedback in the feedback...
Use Try Again, Don't Look Back: Blind Resampling Outperforms Self-Repair in Small Code Models to bound risk before recurring or unattended execution.
Research source arXiv:2607.26117; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,704
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1704
2026-07-30
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.26117
[2607.26117] Try Again, Don't Look Back: Blind Resampling Outperforms Self-Repair in Small Code Models
Self-repair - returning a failed program to the model together with its test output and asking for a correction - is a standard component of code agents, and is almost always evaluated against a baseline that does not retry at all. We argue that this comparison confounds the value of the feedback with the value of the ...
Yuvraj Verma
2026-07-28
2026
arXiv
arXiv
Code, pre-registrations and run traces: https://github.com/vermayuvraj/self-improving-agent
cs.SE
arxiv-api
2607.26117
2026-09-04T05:23:38
ale-0950
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Even More Deception: Objective Misalignment in Mixed-Motive LLM Multi-Agent Systems
https://arxiv.org/abs/2607.26120
external
arxiv.org
In a Werewolf testbed, agents with hidden or conflicting objectives show detectable shifts in internal reasoning while public messages mask the change, degrading collective decisions. Evidence that monitoring inter-agent transcripts alone will not catch misaligned delegates.
In a Werewolf testbed, agents with hidden or conflicting objectives show detectable shifts in internal reasoning while public messages mask the change, degrading collective decisions. Evidence that monitoring inter-agent transcripts alone will not catch misaligned delegates.
In a Werewolf testbed, agents with hidden or conflicting objectives show detectable shifts in internal reasoning while public messages mask the change, degrading collective decisions. Evidence that monitoring inter-agent transcripts alone will not catch misaligned delegates.
The work separates roles across agents, verifiers, or orchestration layers. In a Werewolf testbed, agents with hidden or conflicting objectives show detectable shifts in internal reasoning while public messages mask the change, degrading collective decisions. Evidence that monitoring inter-agent transcripts alone will ...
Use Even More Deception: Objective Misalignment in Mixed-Motive LLM Multi-Agent Systems to bound risk before recurring or unattended execution.
Research source arXiv:2607.26120; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,705
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1705
2026-07-30
Govern
govern
Bound permissions, cost, failure, and escalation.
objective;delegation
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.26120
[2607.26120] Even More Deception: Objective Misalignment in Mixed-Motive LLM Multi-Agent Systems
Large Language Models (LLMs)-powered multi-agent systems are increasingly deployed in mixed-motive environments, where agents operate under asymmetric information and strategic deception due to conflicting or hidden objectives. In these settings, misalignment with collective goals becomes a central concern. We propose ...
Marylou Fauchard; Florian Carichon; Margarida Carvalho; Golnoosh Farnadi
2026-07-28
2026
arXiv
arXiv
Accepted at AIWILD@ICLR 2026
cs.AI
arxiv-api
2607.26120
2026-09-04T05:23:38
ale-0951
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
(Im)Paired Programming: Coding Agents Improve Productivity but Harm Understanding
https://arxiv.org/abs/2607.26375
external
arxiv.org
Controlled study of 54 students comparing an agent-based system against a chatbot: agents raise completion speed but lower code comprehension and the ability to extend the work independently, with copy-paste prompting predicting the weakest understanding. Quantifies the comprehension debt that accrues inside sustained ...
Controlled study of 54 students comparing an agent-based system against a chatbot: agents raise completion speed but lower code comprehension and the ability to extend the work independently, with copy-paste prompting predicting the weakest understanding. Quantifies the comprehension debt that accrues inside sustained ...
Controlled study of 54 students comparing an agent-based system against a chatbot: agents raise completion speed but lower code comprehension and the ability to extend the work independently, with copy-paste prompting predicting the weakest understanding. Quantifies the comprehension debt that accrues inside sustained ...
Keeps adoption grounded in known failure modes, economics, and operational limits. Controlled study of 54 students comparing an agent-based system against a chatbot: agents raise completion speed but lower code comprehension and the ability to extend the work independently, with copy-paste prompting predicting the weak...
Use (Im)Paired Programming: Coding Agents Improve Productivity but Harm Understanding to bound risk before recurring or unattended execution.
Research source arXiv:2607.26375; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,706
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1706
2026-07-30
Govern
govern
Bound permissions, cost, failure, and escalation.
exit
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.26375
[2607.26375] (Im)Paired Programming: Coding Agents Improve Productivity but Harm Understanding
Coding agents (e.g., Cursor) improve developer productivity by optimizing task completion, but shifting users from writing code to prompting and reviewing may harm their understanding, impeding oversight, learning, and communication. To probe this, we have 54 students create a website with one of two AI systems: an age...
Nishant Balepur; Connor Baumler; Valerie Chen; Eunsol Choi; Rachel Rudinger; Jordan Lee Boyd-Graber
2026-07-29
2026
arXiv
arXiv
In-progress Preprint
cs.CL
arxiv-api
2607.26375
2026-09-04T05:23:38
ale-0952
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Blog
📝
We Gave GPT-5.6 Sol a Real Business
https://www.bottlenecklabs.com/blog/autonomously-run-businesses
external
www.bottlenecklabs.com
Field report from handing a frontier model an actual operating business and letting it run unattended, recording where the loop held up and where it needed a human. Useful as evidence about long-horizon autonomy outside benchmark conditions.
Field report from handing a frontier model an actual operating business and letting it run unattended, recording where the loop held up and where it needed a human. Useful as evidence about long-horizon autonomy outside benchmark conditions.
Field report from handing a frontier model an actual operating business and letting it run unattended, recording where the loop held up and where it needed a human. Useful as evidence about long-horizon autonomy outside benchmark conditions.
The work turns loop quality into a measurable task or score. Field report from handing a frontier model an actual operating business and letting it run unattended, recording where the loop held up and where it needed a human. Useful as evidence about long-horizon autonomy outside benchmark conditions.
Use We Gave GPT-5.6 Sol a Real Business to bound risk before recurring or unattended execution.
Contextual source from www.bottlenecklabs.com; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,707
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1707
2026-07-30
Govern
govern
Bound permissions, cost, failure, and escalation.
verification;escalation
operator;security
cross-layer
enabling
practitioner-analysis
B
ok
https://www.bottlenecklabs.com/blog/autonomously-run-businesses
GPT 5.6 Sol Ran a Real Business | Bottleneck Labs
If an agent had a wallet, a computer, and 24 hours, could it run a profitable startup?
Bottleneck Labs
html-meta
2026-09-04T05:23:38
ale-0953
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Blog
📝
GCC Steering Committee Announces AI Policy
https://lwn.net/Articles/1086041/
external
lwn.net
LWN, 2026-07-29. The GCC steering committee will decline any legally significant contribution that includes or is derived from LLM-generated content, with 'legally significant' pegged at roughly 15 lines per GNU maintainer guidelines.
LWN, 2026-07-29. The GCC steering committee will decline any legally significant contribution that includes or is derived from LLM-generated content, with 'legally significant' pegged at roughly 15 lines per GNU maintainer guidelines.
LWN, 2026-07-29. The GCC steering committee will decline any legally significant contribution that includes or is derived from LLM-generated content, with 'legally significant' pegged at roughly 15 lines per GNU maintainer guidelines.
Keeps adoption grounded in known failure modes, economics, and operational limits. LWN, 2026-07-29. The GCC steering committee will decline any legally significant contribution that includes or is derived from LLM-generated content, with 'legally significant' pegged at roughly 15 lines per GNU maintainer guidelines.
Use GCC Steering Committee Announces AI Policy to bound risk before recurring or unattended execution.
Contextual source from lwn.net; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,708
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1708
2026-07-30
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
operator;security
cross-layer
enabling
practitioner-analysis
B
ok
https://lwn.net/Articles/1086041/
GCC steering committee announces AI policy [LWN.net]
The GCC steering committee has announced that it has accepted an AI contributions policy recomm [...]
LWN.net
html-meta
2026-09-04T05:23:38
ale-0954
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Beyond the Leaderboard: A Synthesis of Tool-Use, Planning, and Reasoning Failures in Large Language Model Agents
https://arxiv.org/abs/2607.05775
external
arxiv.org
A cross-cutting synthesis of 27 benchmark, taxonomy and audit papers from 2023-2026 spanning 19 distinct benchmarks, collapsed into six failure clusters: tool invocation and parameter-level errors, planning and constraint-satisfaction failures, long-horizon degradation from context accumulation, multi-agent coordinatio...
A cross-cutting synthesis of 27 benchmark, taxonomy and audit papers from 2023-2026 spanning 19 distinct benchmarks, collapsed into six failure clusters: tool invocation and parameter-level errors, planning and constraint-satisfaction failures, long-horizon degradation from context accumulation, multi-agent coordinatio...
A cross-cutting synthesis of 27 benchmark, taxonomy and audit papers from 2023-2026 spanning 19 distinct benchmarks, collapsed into six failure clusters: tool invocation and parameter-level errors, planning and constraint-satisfaction failures, long-horizon degradation from context accumulation, multi-agent coordinatio...
The work turns loop quality into a measurable task or score. A cross-cutting synthesis of 27 benchmark, taxonomy and audit papers from 2023-2026 spanning 19 distinct benchmarks, collapsed into six failure clusters: tool invocation and parameter-level errors, planning and constraint-satisfaction failures, long-horizon d...
Use Beyond the Leaderboard: A Synthesis of Tool-Use, Planning, and Reasoning Failures in Large Language Model Agents to bound risk before recurring or unattended execution.
Research source arXiv:2607.05775; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,709
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1709
2026-07-30
Govern
govern
Bound permissions, cost, failure, and escalation.
intake;workspace;context;delegation;verification
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.05775
[2607.05775] Beyond the Leaderboard: A Synthesis of Tool-Use, Planning, and Reasoning Failures in Large Language Model Agents
Large language model (LLM) agents are increasingly evaluated on their ability to use tools, plan multi-step tasks, coordinate with other agents, and operate over extended horizons. Reported benchmark gains often obscure recurring failure modes documented across otherwise unrelated evaluation efforts. This paper synthes...
Wael Albayaydh; Rui Zhao; Ivan Flechais
2026-07-07
2026
arXiv
arXiv
16 pages, 3 tables, 1 figure
cs.AI
arxiv-api
2607.05775
2026-09-04T05:23:38
ale-0955
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Guardrails as Scapegoats: Auditing Unfaithful Safety Refusals in Tool-Augmented LLM Agents
https://arxiv.org/abs/2607.19449
external
arxiv.org
A black-box auditing framework for a failure mode that silently corrupts verification in unattended loops: an agent hits a silent infrastructure failure, a tool returning nothing, a timeout, a malformed response, and reports it to the user as a safety refusal rather than as a broken dependency.
A black-box auditing framework for a failure mode that silently corrupts verification in unattended loops: an agent hits a silent infrastructure failure, a tool returning nothing, a timeout, a malformed response, and reports it to the user as a safety refusal rather than as a broken dependency.
A black-box auditing framework for a failure mode that silently corrupts verification in unattended loops: an agent hits a silent infrastructure failure, a tool returning nothing, a timeout, a malformed response, and reports it to the user as a safety refusal rather than as a broken dependency.
Verification is promoted from a final check to a loop-control signal. A black-box auditing framework for a failure mode that silently corrupts verification in unattended loops: an agent hits a silent infrastructure failure, a tool returning nothing, a timeout, a malformed response, and reports it to the user as a safet...
Use Guardrails as Scapegoats: Auditing Unfaithful Safety Refusals in Tool-Augmented LLM Agents to bound risk before recurring or unattended execution.
Research source arXiv:2607.19449; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,710
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1710
2026-07-30
Govern
govern
Bound permissions, cost, failure, and escalation.
workspace;verification;budget
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.19449
[2607.19449] Guardrails as Scapegoats: Auditing Unfaithful Safety Refusals in Tool-Augmented LLM Agents
Evaluation frameworks for tool-augmented LLM agents focus overwhelmingly on capability metrics or explicit tool crashes, leaving silent infrastructure failures and HTTP 200 responses with empty, null, or malformed payloads largely unaudited. We introduce a lightweight black-box auditing framework that injects four sile...
Aarushi Singh
2026-07-21
2026
arXiv
arXiv
10 pages, 3 figures. Accepted at the ACM KDD 2026 Workshop on Evaluation and Trustworthiness of Agentic AI
cs.LG
arxiv-api
2607.19449
2026-09-04T05:23:38
ale-0956
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Towards a Science of AI Agent Reliability
https://arxiv.org/abs/2602.16666
external
arxiv.org
The foundational negative result behind the reliability-versus-capability argument that several entries in this list gesture at without citing.
The foundational negative result behind the reliability-versus-capability argument that several entries in this list gesture at without citing.
The foundational negative result behind the reliability-versus-capability argument that several entries in this list gesture at without citing.
Keeps adoption grounded in known failure modes, economics, and operational limits. The foundational negative result behind the reliability-versus-capability argument that several entries in this list gesture at without citing.
Use Towards a Science of AI Agent Reliability to bound risk before recurring or unattended execution.
Research source arXiv:2602.16666; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,711
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1711
2026-07-30
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2602.16666
[2602.16666] Towards a Science of AI Agent Reliability
AI agents are increasingly deployed to execute important tasks. While rising accuracy scores on standard benchmarks suggest rapid progress, many agents still continue to fail in practice. This discrepancy highlights a fundamental limitation of current evaluations: compressing agent behavior into a single success metric...
Stephan Rabanser; Sayash Kapoor; Peter Kirgis; Kangheng Liu; Saiteja Utpala; Arvind Narayanan
2026-02-18
2026
arXiv
arXiv
Accepted at ICML 2026. Interactive dashboard available at: https://hal.cs.princeton.edu/reliability
cs.AI
arxiv-api
2602.16666
2026-09-04T05:23:38
ale-0957
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Blog
📝
Ten AI Agents Destroyed Production, Zero Postmortems
https://www.harperfoley.com/blog/ai-agents-destroyed-production-zero-postmortems
external
www.harperfoley.com
Harper Foley (Tribe AI, ex-Navy EOD) catalogs ten production-destroying agent incidents across six tools over 16 months, each sourced to GitHub issues, Fortune, The Register, or first-hand reports, and shows not one vendor published a postmortem.
Harper Foley (Tribe AI, ex-Navy EOD) catalogs ten production-destroying agent incidents across six tools over 16 months, each sourced to GitHub issues, Fortune, The Register, or first-hand reports, and shows not one vendor published a postmortem.
Harper Foley (Tribe AI, ex-Navy EOD) catalogs ten production-destroying agent incidents across six tools over 16 months, each sourced to GitHub issues, Fortune, The Register, or first-hand reports, and shows not one vendor published a postmortem.
Keeps adoption grounded in known failure modes, economics, and operational limits. Harper Foley (Tribe AI, ex-Navy EOD) catalogs ten production-destroying agent incidents across six tools over 16 months, each sourced to GitHub issues, Fortune, The Register, or first-hand reports, and shows not one vendor published a po...
Use Ten AI Agents Destroyed Production, Zero Postmortems to bound risk before recurring or unattended execution.
Contextual source from www.harperfoley.com; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,712
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1712
2026-08-02
Govern
govern
Bound permissions, cost, failure, and escalation.
intake;workspace
operator;security
cross-layer
enabling
practitioner-analysis
B
ok
https://www.harperfoley.com/blog/ai-agents-destroyed-production-zero-postmortems
Ten AI Agents Destroyed Production. Zero Postmortems. | Harper Foley
10 documented incidents across 6 AI coding tools in 16 months. Missing audit trails, no liability frameworks, no vendor postmortems. The accountability infrastructure doesn't exist.
Harper Foley
2026-03-08
2026
Harper Foley - AI Product Leader
html-meta
2026-09-04T05:23:38
ale-0958
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Blog
📝
2x, Not 10x: Coding With LLMs in 2026
https://obryant.dev/p/2x-not-10x/
external
obryant.dev
A calibration essay whose central claim is squarely a loop-engineering claim: LLMs became genuinely useful at the point they got reliable enough to run inside automated feedback loops, and past that threshold further model capability buys much less than retooling does.
A calibration essay whose central claim is squarely a loop-engineering claim: LLMs became genuinely useful at the point they got reliable enough to run inside automated feedback loops, and past that threshold further model capability buys much less than retooling does.
A calibration essay whose central claim is squarely a loop-engineering claim: LLMs became genuinely useful at the point they got reliable enough to run inside automated feedback loops, and past that threshold further model capability buys much less than retooling does.
Keeps adoption grounded in known failure modes, economics, and operational limits. A calibration essay whose central claim is squarely a loop-engineering claim: LLMs became genuinely useful at the point they got reliable enough to run inside automated feedback loops, and past that threshold further model capability buy...
Use 2x, Not 10x: Coding With LLMs in 2026 to bound risk before recurring or unattended execution.
Contextual source from obryant.dev; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,713
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1713
2026-08-02
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
operator;security
cross-layer
enabling
practitioner-analysis
B
ok
https://obryant.dev/p/2x-not-10x/
2x, not 10x: coding with LLMs in 2026
Calibrate your enthusiasm
Jacob O'Bryant
obryant.dev
html-meta
2026-09-04T05:23:38
ale-0959
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Reproducing LightMem: Naive RAG Is Just as Good for Memory Management
https://arxiv.org/abs/2607.29104
external
arxiv.org
A reproduction study that lands harder than most original results. The authors rebuild LightMem, a well-cited lightweight memory-management approach, and compare it against naive RAG retrieving directly from raw user turns.
A reproduction study that lands harder than most original results. The authors rebuild LightMem, a well-cited lightweight memory-management approach, and compare it against naive RAG retrieving directly from raw user turns.
A reproduction study that lands harder than most original results. The authors rebuild LightMem, a well-cited lightweight memory-management approach, and compare it against naive RAG retrieving directly from raw user turns.
Persistent memory is treated as an external runtime artifact. A reproduction study that lands harder than most original results. The authors rebuild LightMem, a well-cited lightweight memory-management approach, and compare it against naive RAG retrieving directly from raw user turns.
Use Reproducing LightMem: Naive RAG Is Just as Good for Memory Management to bound risk before recurring or unattended execution.
Research source arXiv:2607.29104; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,714
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1714
2026-08-05
Govern
govern
Bound permissions, cost, failure, and escalation.
context
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.29104
[2607.29104] Reproducing LightMem: Naive RAG Is Just as Good for Memory Management
Long-term conversational agents require access to information from earlier interactions, such as a user's preferences, past requests, or previously mentioned facts. Repeatedly providing the full dialogue history can be expensive as conversations grow, so many memory approaches instead transform past interactions into c...
Yongjie Zhou; Shuai Wang; Bevan Koopman; Guido Zuccon
2026-07-31
2026
arXiv
arXiv
Code: https://github.com/ielab/Reproducing-LightMem
cs.IR
arxiv-api
2607.29104
2026-09-04T05:23:38
ale-0960
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Blog
📝
Critical CVEs or LLM Slops? Inside 50+ Fake SQLite Advisories
https://research.jfrog.com/post/sqlite-critical-cves-or-llm-slops/
external
research.jfrog.com
JFrog Security Research (Afek Berger) audited 55 SQLite vulnerability advisories published from a single GitHub account with initial CVSS scores of 7.5 to 9.8, and found 54 completely fabricated, non-existent functions, invalid line numbers, PoCs that trigger no crash, with one real bug wrapped in unverified CVE metada...
JFrog Security Research (Afek Berger) audited 55 SQLite vulnerability advisories published from a single GitHub account with initial CVSS scores of 7.5 to 9.8, and found 54 completely fabricated, non-existent functions, invalid line numbers, PoCs that trigger no crash, with one real bug wrapped in unverified CVE metada...
JFrog Security Research (Afek Berger) audited 55 SQLite vulnerability advisories published from a single GitHub account with initial CVSS scores of 7.5 to 9.8, and found 54 completely fabricated, non-existent functions, invalid line numbers, PoCs that trigger no crash, with one real bug wrapped in unverified CVE metada...
Keeps adoption grounded in known failure modes, economics, and operational limits. JFrog Security Research (Afek Berger) audited 55 SQLite vulnerability advisories published from a single GitHub account with initial CVSS scores of 7.5 to 9.8, and found 54 completely fabricated, non-existent functions, invalid line numb...
Use Critical CVEs or LLM Slops? Inside 50+ Fake SQLite Advisories to bound risk before recurring or unattended execution.
Contextual source from research.jfrog.com; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,715
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1715
2026-08-05
Govern
govern
Bound permissions, cost, failure, and escalation.
trigger
operator;security
cross-layer
enabling
practitioner-analysis
B
ok
https://research.jfrog.com/post/sqlite-critical-cves-or-llm-slops/
SQLite Critical CVEs or LLM Slop? - JFrog Security Research
The JFrog security research team recently identified a supply chain attack targeting the `xinference` package on PyPI. Versions 2.6.0, 2.6.1, and 2.6.2 were compromised and yanked by maintainers after users reported suspicious behavior. If you installed or imported these versions, you must assume your environment is co...
research.jfrog.com
domain-fallback
2026-09-04T05:23:38
ale-0961
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Blog
📝
Don't be a meat proxy
https://gruhn.me/blog/2026-08-03/
external
gruhn.me
The highest-traction technical post of the Aug 1-3 window, arguing against the degenerate role humans fall into when they relay model output verbatim, into Slack threads, PR review comments, group chats, without reading, understanding or validating it first.
The highest-traction technical post of the Aug 1-3 window, arguing against the degenerate role humans fall into when they relay model output verbatim, into Slack threads, PR review comments, group chats, without reading, understanding or validating it first.
The highest-traction technical post of the Aug 1-3 window, arguing against the degenerate role humans fall into when they relay model output verbatim, into Slack threads, PR review comments, group chats, without reading, understanding or validating it first.
Keeps adoption grounded in known failure modes, economics, and operational limits. The highest-traction technical post of the Aug 1-3 window, arguing against the degenerate role humans fall into when they relay model output verbatim, into Slack threads, PR review comments, group chats, without reading, understanding or...
Use Don't be a meat proxy to bound risk before recurring or unattended execution.
Contextual source from gruhn.me; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,716
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1716
2026-08-05
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
operator;security
cross-layer
enabling
practitioner-analysis
B
ok
https://gruhn.me/blog/2026-08-03/
Niklas Gruhn - Don't be a meat proxy
2026
gruhn.me
url-date
2026-09-04T05:23:38
ale-0962
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
What Breaks When LLMs Code? Characterizing Operational Safety Failures of Agentic Code Assistants
https://arxiv.org/abs/2605.30777
external
arxiv.org
An incident-driven empirical study rather than a benchmark paper: the authors screened 68,816 papers across 22 venues to curate 185 safety-relevant studies, then mined 16,586 GitHub issues from LLM-powered coding tools and manually confirmed 547 genuine operational safety failures, each annotated with contributing fact...
An incident-driven empirical study rather than a benchmark paper: the authors screened 68,816 papers across 22 venues to curate 185 safety-relevant studies, then mined 16,586 GitHub issues from LLM-powered coding tools and manually confirmed 547 genuine operational safety failures, each annotated with contributing fact...
An incident-driven empirical study rather than a benchmark paper: the authors screened 68,816 papers across 22 venues to curate 185 safety-relevant studies, then mined 16,586 GitHub issues from LLM-powered coding tools and manually confirmed 547 genuine operational safety failures, each annotated with contributing fact...
The work turns loop quality into a measurable task or score. An incident-driven empirical study rather than a benchmark paper: the authors screened 68,816 papers across 22 venues to curate 185 safety-relevant studies, then mined 16,586 GitHub issues from LLM-powered coding tools and manually confirmed 547 genuine opera...
Use What Breaks When LLMs Code? Characterizing Operational Safety Failures of Agentic Code Assistants to bound risk before recurring or unattended execution.
Research source arXiv:2605.30777; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,717
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1717
2026-08-05
Govern
govern
Bound permissions, cost, failure, and escalation.
intake;workspace;context;verification
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2605.30777
[2605.30777] What Breaks When LLMs Code? Characterizing Operational Safety Failures of Agentic Code Assistants
Autonomous coding agents built on large language models (LLMs) are rapidly being integrated into development workflows, yet their operational safety properties remain poorly understood beyond evaluations of explicitly malicious inputs. In practice, high-impact failures arise during benign, goal-directed use through env...
Alif Al Hasan; Sumon Biswas
2026-05-29
2026
arXiv
arXiv
10.1145/3832783.3834393
This paper is accepted to the 41st IEEE/ACM International Conference on Automated Software Engineering (ASE 2026), Research Track
cs.SE
arxiv-api
2605.30777
2026-09-04T05:23:38
ale-0963
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
How Coding Agents Fail Their Users: A Large-Scale Analysis of Developer-Agent Misalignment in 20,574 Real-World Sessions
https://arxiv.org/abs/2605.29442
external
arxiv.org
A study of 20,574 real coding-agent sessions across 1,639 repositories, categorizing seven recurring forms of developer-agent misalignment and separating IDE from CLI usage patterns.
A study of 20,574 real coding-agent sessions across 1,639 repositories, categorizing seven recurring forms of developer-agent misalignment and separating IDE from CLI usage patterns.
A study of 20,574 real coding-agent sessions across 1,639 repositories, categorizing seven recurring forms of developer-agent misalignment and separating IDE from CLI usage patterns.
Keeps adoption grounded in known failure modes, economics, and operational limits. A study of 20,574 real coding-agent sessions across 1,639 repositories, categorizing seven recurring forms of developer-agent misalignment and separating IDE from CLI usage patterns.
Use How Coding Agents Fail Their Users: A Large-Scale Analysis of Developer-Agent Misalignment in 20,574 Real-World Sessions to bound risk before recurring or unattended execution.
Research source arXiv:2605.29442; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,718
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1718
2026-08-05
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2605.29442
[2605.29442] How Coding Agents Fail Their Users: A Large-Scale Analysis of Developer-Agent Misalignment in 20,574 Real-World Sessions
AI coding agents increasingly act directly within software environments, yet existing analyses of their failures rely on benchmark trajectories that miss how developers actually experience misalignment. We present an observational study of 20,574 coding-agent sessions from 1,639 repositories across IDE and CLI workflow...
Ningzhi Tang; Chaoran Chen; Gelei Xu; Yiyu Shi; Yu Huang; Collin McMillan; Tao Dong; Toby Jia-Jun Li
2026-05-28
2026
arXiv
arXiv
cs.SE
arxiv-api
2605.29442
2026-09-04T05:23:38
ale-0964
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Skill Use or Skill Theater? Evaluating the Reasoning Backroom in Skill-Augmented Language Agents
https://arxiv.org/abs/2607.27484
external
arxiv.org
BACKTRACE measures whether a skill actually caused a decision by comparing skill-conditioned against no-skill runs and perturbing skill attributes.
BACKTRACE measures whether a skill actually caused a decision by comparing skill-conditioned against no-skill runs and perturbing skill attributes.
BACKTRACE measures whether a skill actually caused a decision by comparing skill-conditioned against no-skill runs and perturbing skill attributes.
Keeps adoption grounded in known failure modes, economics, and operational limits. BACKTRACE measures whether a skill actually caused a decision by comparing skill-conditioned against no-skill runs and perturbing skill attributes.
Use Skill Use or Skill Theater? Evaluating the Reasoning Backroom in Skill-Augmented Language Agents to bound risk before recurring or unattended execution.
Research source arXiv:2607.27484; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,719
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1719
2026-08-05
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.27484
[2607.27484] Skill Use or Skill Theater? Evaluating the Reasoning Backroom in Skill-Augmented Language Agents
Reusable skills are becoming a standard interface for extending language agents with task procedures. Yet evaluators usually infer skill use from visible reasoning or the agent's own attribution. These signals show what the agent appears to use, not whether the skill changed its decision. We ask whether skill-augmented...
Jinwei Hu; Yi Qi; Xinmiao Huang; Youcheng Sun; Yi Dong; Xiaowei Huang
2026-07-29
2026
arXiv
arXiv
21 pages
cs.AI
arxiv-api
2607.27484
2026-09-04T05:23:38
ale-0965
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Fidelity Is Not Safety: Gently-Compressed LLMs Pass Every Data-Free Quality Guard Yet Invent Procedure Steps in Agentic Execution
https://arxiv.org/abs/2607.28196
external
arxiv.org
Compressed models clear perplexity, downstream accuracy, and data-free output-fidelity checks, then fabricate procedure steps once deployed in an agent loop, and the effect is specific to coherent low-rank (SVD) error, not magnitude pruning at matched perplexity.
Compressed models clear perplexity, downstream accuracy, and data-free output-fidelity checks, then fabricate procedure steps once deployed in an agent loop, and the effect is specific to coherent low-rank (SVD) error, not magnitude pruning at matched perplexity.
Compressed models clear perplexity, downstream accuracy, and data-free output-fidelity checks, then fabricate procedure steps once deployed in an agent loop, and the effect is specific to coherent low-rank (SVD) error, not magnitude pruning at matched perplexity.
Keeps adoption grounded in known failure modes, economics, and operational limits. Compressed models clear perplexity, downstream accuracy, and data-free output-fidelity checks, then fabricate procedure steps once deployed in an agent loop, and the effect is specific to coherent low-rank (SVD) error, not magnitude prun...
Use Fidelity Is Not Safety: Gently-Compressed LLMs Pass Every Data-Free Quality Guard Yet Invent Procedure Steps in Agentic Execution to bound risk before recurring or unattended execution.
Research source arXiv:2607.28196; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,720
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1720
2026-08-05
Govern
govern
Bound permissions, cost, failure, and escalation.
budget;escalation;exit
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.28196
[2607.28196] Fidelity Is Not Safety: Gently-Compressed LLMs Pass Every Data-Free Quality Guard Yet Invent Procedure Steps in Agentic Execution
Practitioners accept a compressed language model once it clears a stack of data-cheap quality guards: perplexity within a small factor of the original, downstream accuracy (for example MMLU) inside a confidence interval, and data-free output-fidelity signals that compare the compressed and original network's internal r...
I. Kennedy; T. Kennedy
2026-07-30
2026
arXiv
arXiv
cs.CL
arxiv-api
2607.28196
2026-09-04T05:23:38
ale-0966
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Rethinking Inference-Time Scaling in Local Computer-Use Agents: Failure Modes and Compute Tradeoffs
https://arxiv.org/abs/2607.28573
external
arxiv.org
Tests whether spending more inference compute rescues small locally-hosted computer-use agents on OSWorld, and finds it mostly changes the failure mode rather than fixing it: contextual scaling stabilizes trajectories, but temporal scaling extends wrong paths instead of correcting them. A useful corrective to 'just let...
Tests whether spending more inference compute rescues small locally-hosted computer-use agents on OSWorld, and finds it mostly changes the failure mode rather than fixing it: contextual scaling stabilizes trajectories, but temporal scaling extends wrong paths instead of correcting them. A useful corrective to 'just let...
Tests whether spending more inference compute rescues small locally-hosted computer-use agents on OSWorld, and finds it mostly changes the failure mode rather than fixing it: contextual scaling stabilizes trajectories, but temporal scaling extends wrong paths instead of correcting them. A useful corrective to 'just let...
Keeps adoption grounded in known failure modes, economics, and operational limits. Tests whether spending more inference compute rescues small locally-hosted computer-use agents on OSWorld, and finds it mostly changes the failure mode rather than fixing it: contextual scaling stabilizes trajectories, but temporal scali...
Use Rethinking Inference-Time Scaling in Local Computer-Use Agents: Failure Modes and Compute Tradeoffs to bound risk before recurring or unattended execution.
Research source arXiv:2607.28573; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,721
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1721
2026-08-05
Govern
govern
Bound permissions, cost, failure, and escalation.
verification
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.28573
[2607.28573] Rethinking Inference-Time Scaling in Local Computer-Use Agents: Failure Modes and Compute Tradeoffs
Deploying autonomous computer-use agents (CUAs) locally is increasingly important for privacy, cost efficiency, and practical usability, yet improving their performance under strict hardware constraints remains challenging. While recent studies show that inference-time scaling can improve frontier computer-use agents t...
Woongkyu Lee; Jungwook Choi
2026-07-30
2026
arXiv
arXiv
cs.AI
arxiv-api
2607.28573
2026-09-04T05:23:38
ale-0967
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Safeguards Based on Copyable Context Cannot Provide Reliable Safety for LLMs
https://arxiv.org/abs/2607.27951
external
arxiv.org
Submitted 2026-07-30. An impossibility argument aimed squarely at the standard loop-safety design: if the evidence a safeguard uses (system prompt, conversation history, stated role) is copyable, an attacker can imitate it.
Submitted 2026-07-30. An impossibility argument aimed squarely at the standard loop-safety design: if the evidence a safeguard uses (system prompt, conversation history, stated role) is copyable, an attacker can imitate it.
Submitted 2026-07-30. An impossibility argument aimed squarely at the standard loop-safety design: if the evidence a safeguard uses (system prompt, conversation history, stated role) is copyable, an attacker can imitate it.
Context is managed as durable loop state rather than a single prompt payload. Submitted 2026-07-30. An impossibility argument aimed squarely at the standard loop-safety design: if the evidence a safeguard uses (system prompt, conversation history, stated role) is copyable, an attacker can imitate it.
Use Safeguards Based on Copyable Context Cannot Provide Reliable Safety for LLMs to bound risk before recurring or unattended execution.
Research source arXiv:2607.27951; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,722
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1722
2026-08-07
Govern
govern
Bound permissions, cost, failure, and escalation.
context
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.27951
[2607.27951] Safeguards Based on Copyable Context Cannot Provide Reliable Safety for LLMs
Large language model safeguards decide whether to answer before seeing how an answer will be used. This creates a basic problem for dual-use tasks: the same answer can help an authorized professional or an attacker, while an attacker can imitate a benign request and interaction history. We separate the capability relea...
Pingyu Wu; Lingyao Zhu; Weiming Zhang; Nenghai Yu
2026-07-30
2026
arXiv
arXiv
cs.CR
arxiv-api
2607.27951
2026-09-04T05:23:38
ale-0968
Critiques, Risks, And Limitations
critiques-risks-and-limitations
Paper
📄
Do Context Files Help Coding Agents? A Two-Agent Ablation Study on Real Repositories
https://arxiv.org/abs/2607.27250
external
arxiv.org
Submitted 2026-07-28. Controlled ablation of AGENTS.md / CLAUDE.md context files across two frontier agents, 17 real repository tasks, and 288 evaluated runs. Finding: context strategy does not measurably move correctness on either agent, and failures are dominated by implementation difficulty rather than missing repo ...
Submitted 2026-07-28. Controlled ablation of AGENTS.md / CLAUDE.md context files across two frontier agents, 17 real repository tasks, and 288 evaluated runs. Finding: context strategy does not measurably move correctness on either agent, and failures are dominated by implementation difficulty rather than missing repo ...
Submitted 2026-07-28. Controlled ablation of AGENTS.md / CLAUDE.md context files across two frontier agents, 17 real repository tasks, and 288 evaluated runs. Finding: context strategy does not measurably move correctness on either agent, and failures are dominated by implementation difficulty rather than missing repo ...
Context is managed as durable loop state rather than a single prompt payload. Submitted 2026-07-28. Controlled ablation of AGENTS.md / CLAUDE.md context files across two frontier agents, 17 real repository tasks, and 288 evaluated runs. Finding: context strategy does not measurably move correctness on either agent, and...
Use Do Context Files Help Coding Agents? A Two-Agent Ablation Study on Real Repositories to bound risk before recurring or unattended execution.
Research source arXiv:2607.27250; inspect its method and evaluation before treating results as production evidence.
medium
README.md
1,723
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1723
2026-08-07
Govern
govern
Bound permissions, cost, failure, and escalation.
context
researcher;evaluator;operator;security
cross-layer
enabling
research-preprint
A
ok
https://arxiv.org/abs/2607.27250
[2607.27250] Do Context Files Help Coding Agents? A Two-Agent Ablation Study on Real Repositories
Persistent context files (AGENTS.md, CLAUDE.md) are standard practice for guiding AI coding agents, yet evidence for their effectiveness is contradictory. We present a controlled ablation of context-injection strategy across two frontier agents (Claude Code and Codex), 17 real tasks from 3 repositories (15 shared + 2 C...
Prakhar Khatri
2026-07-28
2026
arXiv
arXiv
cs.SE
arxiv-api
2607.27250
2026-09-04T05:23:38
ale-0969
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
Awesome Harness Engineering by ai-boost
https://github.com/ai-boost/awesome-harness-engineering
external
github.com
Comprehensive list for the agent harness layer that Loop Engineering builds on.
Comprehensive list for the agent harness layer that Loop Engineering builds on.
Comprehensive list for the agent harness layer that Loop Engineering builds on.
Connects neighboring ecosystems while preserving Loop Engineering as a narrower operating concept. Comprehensive list for the agent harness layer that Loop Engineering builds on.
Use Awesome Harness Engineering by ai-boost to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,743
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1743
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder
cross-layer
adjacent
curated-index
C
ok
https://github.com/ai-boost/awesome-harness-engineering
GitHub - ai-boost/awesome-harness-engineering: Awesome list for AI agent harness engineering: tools, patterns, evals, memory, MCP, permissions, observability, and orchestration. · GitHub
Awesome list for AI agent harness engineering: tools, patterns, evals, memory, MCP, permissions, observability, and orchestration. - ai-boost/awesome-harness-engineering
ai-boost/awesome-harness-engineering
GitHub
html-meta
ai-boost/awesome-harness-engineering
2026-09-04T05:23:38
ale-0970
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
Awesome Harness Engineering by walkinglabs
https://github.com/walkinglabs/awesome-harness-engineering
external
github.com
High-signal harness list with strong categories for context, guardrails, specs, evals, runtimes, and benchmarks.
High-signal harness list with strong categories for context, guardrails, specs, evals, runtimes, and benchmarks.
High-signal harness list with strong categories for context, guardrails, specs, evals, runtimes, and benchmarks.
Evaluation data is used as the feedback signal for improving loop behavior. High-signal harness list with strong categories for context, guardrails, specs, evals, runtimes, and benchmarks.
Use Awesome Harness Engineering by walkinglabs to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,744
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1744
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
context;verification
builder
cross-layer
adjacent
curated-index
C
ok
https://github.com/walkinglabs/awesome-harness-engineering
GitHub - walkinglabs/awesome-harness-engineering: 🛠️ Awesome tools & guides for harness engineering. · GitHub
🛠️ Awesome tools & guides for harness engineering. - walkinglabs/awesome-harness-engineering
walkinglabs/awesome-harness-engineering
GitHub
html-meta
walkinglabs/awesome-harness-engineering
2026-09-04T05:23:38
ale-0971
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
Awesome Agent Harness
https://github.com/AutoJunjie/awesome-agent-harness
external
github.com
Curated tools and resources for environments, constraints, and feedback around coding agents.
Curated tools and resources for environments, constraints, and feedback around coding agents.
Curated tools and resources for environments, constraints, and feedback around coding agents.
Connects neighboring ecosystems while preserving Loop Engineering as a narrower operating concept. Curated tools and resources for environments, constraints, and feedback around coding agents.
Use Awesome Agent Harness to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,745
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1745
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
workspace
builder
cross-layer
adjacent
curated-index
C
ok
https://github.com/AutoJunjie/awesome-agent-harness
GitHub - AutoJunjie/awesome-agent-harness · GitHub
Contribute to AutoJunjie/awesome-agent-harness development by creating an account on GitHub.
AutoJunjie/awesome-agent-harness
GitHub
html-meta
AutoJunjie/awesome-agent-harness
2026-09-04T05:23:38
ale-0972
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
Awesome Context Engineering
https://github.com/Meirtz/Awesome-Context-Engineering
external
github.com
Survey-style list for context engineering across LLMs and agents.
Survey-style list for context engineering across LLMs and agents.
Survey-style list for context engineering across LLMs and agents.
Context is managed as durable loop state rather than a single prompt payload. Survey-style list for context engineering across LLMs and agents.
Use Awesome Context Engineering to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,746
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1746
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
context
builder
cross-layer
adjacent
curated-index
C
ok
https://github.com/Meirtz/Awesome-Context-Engineering
GitHub - Meirtz/Awesome-Context-Engineering: 🔥 Comprehensive survey on Context Engineering: from prompt engineering to production-grade AI systems. hundreds of papers, frameworks, and implementation guides for LLMs and AI agents. · GitHub
🔥 Comprehensive survey on Context Engineering: from prompt engineering to production-grade AI systems. hundreds of papers, frameworks, and implementation guides for LLMs and AI agents. - Meirtz/Awesome-Context-Engineering
Meirtz/Awesome-Context-Engineering
GitHub
html-meta
Meirtz/Awesome-Context-Engineering
2026-09-04T05:23:38
ale-0973
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
Awesome Prompt Engineering
https://github.com/promptslab/Awesome-Prompt-Engineering
external
github.com
Classic adjacent list for prompt techniques and prompting resources.
Classic adjacent list for prompt techniques and prompting resources.
Classic adjacent list for prompt techniques and prompting resources.
Connects neighboring ecosystems while preserving Loop Engineering as a narrower operating concept. Classic adjacent list for prompt techniques and prompting resources.
Use Awesome Prompt Engineering to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,747
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1747
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder
cross-layer
adjacent
curated-index
C
ok
https://github.com/promptslab/Awesome-Prompt-Engineering
GitHub - promptslab/Awesome-Prompt-Engineering: This repository contains a hand-curated resources for Prompt Engineering with a focus on Generative Pre-trained Transformer (GPT), ChatGPT, PaLM etc · GitHub
This repository contains a hand-curated resources for Prompt Engineering with a focus on Generative Pre-trained Transformer (GPT), ChatGPT, PaLM etc - GitHub - promptslab/Awesome-Prompt-Engineering: This repository contains a hand-curated resources for Prompt Engineering with a focus on Generative Pre-trained Transform...
promptslab/Awesome-Prompt-Engineering
GitHub
html-meta
promptslab/Awesome-Prompt-Engineering
2026-09-04T05:23:38
ale-0974
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
Awesome LLM Agents
https://github.com/kaushikb11/awesome-llm-agents
external
github.com
General list of LLM agent papers, frameworks, and applications.
General list of LLM agent papers, frameworks, and applications.
General list of LLM agent papers, frameworks, and applications.
Connects neighboring ecosystems while preserving Loop Engineering as a narrower operating concept. General list of LLM agent papers, frameworks, and applications.
Use Awesome LLM Agents to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,748
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1748
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder
cross-layer
adjacent
curated-index
C
ok
https://github.com/kaushikb11/awesome-llm-agents
GitHub - kaushikb11/awesome-llm-agents: A curated list of awesome LLM agents frameworks. · GitHub
A curated list of awesome LLM agents frameworks. Contribute to kaushikb11/awesome-llm-agents development by creating an account on GitHub.
kaushikb11/awesome-llm-agents
GitHub
html-meta
kaushikb11/awesome-llm-agents
2026-09-04T05:23:38
ale-0975
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
Awesome AI Agents
https://github.com/e2b-dev/awesome-ai-agents
external
github.com
Broad AI agent ecosystem map.
Broad AI agent ecosystem map.
Broad AI agent ecosystem map.
Connects neighboring ecosystems while preserving Loop Engineering as a narrower operating concept. Broad AI agent ecosystem map.
Use Awesome AI Agents to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,749
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1749
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder
cross-layer
adjacent
curated-index
C
ok
https://github.com/e2b-dev/awesome-ai-agents
GitHub - e2b-dev/awesome-ai-agents: A list of AI autonomous agents · GitHub
A list of AI autonomous agents. Contribute to e2b-dev/awesome-ai-agents development by creating an account on GitHub.
e2b-dev/awesome-ai-agents
GitHub
html-meta
e2b-dev/awesome-ai-agents
2026-09-04T05:23:38
ale-0976
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
Awesome CLI Coding Agents
https://github.com/bradAGI/awesome-cli-coding-agents
external
github.com
Directory of terminal-native coding agents, parallel runners, autonomous loops, and the harnesses that orchestrate them.
Directory of terminal-native coding agents, parallel runners, autonomous loops, and the harnesses that orchestrate them.
Directory of terminal-native coding agents, parallel runners, autonomous loops, and the harnesses that orchestrate them.
Orchestration and control flow are made explicit and inspectable. Directory of terminal-native coding agents, parallel runners, autonomous loops, and the harnesses that orchestrate them.
Use Awesome CLI Coding Agents to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,750
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1750
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
delegation
builder
cross-layer
adjacent
curated-index
C
ok
https://github.com/bradAGI/awesome-cli-coding-agents
GitHub - bradAGI/awesome-cli-coding-agents: Curated directory of terminal-native AI coding agents and the harnesses that orchestrate them. Covers open-source tools (Pi, OpenCode, Aider, Goose), platform agents (Claude Code, Codex, Gemini CLI), parallel runners, autonomous loops, and agent infrastructure. · GitHub
Curated directory of terminal-native AI coding agents and the harnesses that orchestrate them. Covers open-source tools (Pi, OpenCode, Aider, Goose), platform agents (Claude Code, Codex, Gemini CLI), parallel runners, autonomous loops, and agent infrastructure. - GitHub - bradAGI/awesome-cli-coding-agents: Curated dire...
bradAGI/awesome-cli-coding-agents
GitHub
html-meta
bradAGI/awesome-cli-coding-agents
2026-09-04T05:23:38
ale-0977
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
Awesome Self-Evolving Agents
https://github.com/XMUDeepLIT/Awesome-Self-Evolving-Agents
external
github.com
Survey-style list of agents that improve themselves over repeated runs, an adjacent angle on long-running loops with memory and verification.
Survey-style list of agents that improve themselves over repeated runs, an adjacent angle on long-running loops with memory and verification.
Survey-style list of agents that improve themselves over repeated runs, an adjacent angle on long-running loops with memory and verification.
Verification is promoted from a final check to a loop-control signal. Survey-style list of agents that improve themselves over repeated runs, an adjacent angle on long-running loops with memory and verification.
Use Awesome Self-Evolving Agents to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,751
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1751
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
context;verification
builder
cross-layer
adjacent
curated-index
C
ok
https://github.com/XMUDeepLIT/Awesome-Self-Evolving-Agents
GitHub - XMUDeepLIT/Awesome-Self-Evolving-Agents: A Survey of Self-Evolving Agents | A curated list of resources (surveys, papers, benchmarks, and opensource projects) on Self-Evolving Agents. · GitHub
A Survey of Self-Evolving Agents | A curated list of resources (surveys, papers, benchmarks, and opensource projects) on Self-Evolving Agents. - XMUDeepLIT/Awesome-Self-Evolving-Agents
XMUDeepLIT/Awesome-Self-Evolving-Agents
GitHub
html-meta
XMUDeepLIT/Awesome-Self-Evolving-Agents
2026-09-04T05:23:38
ale-0978
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
Awesome AI Agent Papers
https://github.com/VoltAgent/awesome-ai-agent-papers
external
github.com
Curated 2026 research collection across agent engineering, memory, evaluation, workflows, and autonomous systems, a paper-level feeder for loop-design foundations.
Curated 2026 research collection across agent engineering, memory, evaluation, workflows, and autonomous systems, a paper-level feeder for loop-design foundations.
Curated 2026 research collection across agent engineering, memory, evaluation, workflows, and autonomous systems, a paper-level feeder for loop-design foundations.
Evaluation data is used as the feedback signal for improving loop behavior. Curated 2026 research collection across agent engineering, memory, evaluation, workflows, and autonomous systems, a paper-level feeder for loop-design foundations.
Use Awesome AI Agent Papers to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,752
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1752
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
context;verification
builder
cross-layer
adjacent
curated-index
C
ok
https://github.com/VoltAgent/awesome-ai-agent-papers
GitHub - VoltAgent/awesome-ai-agent-papers: A curated collection of AI agent research papers released in 2026, covering agent engineering, memory, evaluation, workflows, and autonomous systems. · GitHub
A curated collection of AI agent research papers released in 2026, covering agent engineering, memory, evaluation, workflows, and autonomous systems. - VoltAgent/awesome-ai-agent-papers
VoltAgent/awesome-ai-agent-papers
GitHub
html-meta
VoltAgent/awesome-ai-agent-papers
2026-09-04T05:23:38
ale-0979
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
awesome-ralph
https://github.com/snwfdhmp/awesome-ralph
external
github.com
Curated directory for the Ralph technique, collecting official resources, implementations, playbooks, tutorials, and community channels for running coding agents in automated loops until specifications are fulfilled.
Curated directory for the Ralph technique, collecting official resources, implementations, playbooks, tutorials, and community channels for running coding agents in automated loops until specifications are fulfilled.
Curated directory for the Ralph technique, collecting official resources, implementations, playbooks, tutorials, and community channels for running coding agents in automated loops until specifications are fulfilled.
Primary-source operational guidance rather than commentary. Curated directory for the Ralph technique, collecting official resources, implementations, playbooks, tutorials, and community channels for running coding agents in automated loops until specifications are fulfilled.
Use awesome-ralph to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,753
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1753
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder
cross-layer
adjacent
curated-index
C
ok
https://github.com/snwfdhmp/awesome-ralph
GitHub - snwfdhmp/awesome-ralph: A curated list of resources about Ralph, the AI coding technique that runs AI coding agents in automated loops until specifications are fulfilled. · GitHub
A curated list of resources about Ralph, the AI coding technique that runs AI coding agents in automated loops until specifications are fulfilled. - snwfdhmp/awesome-ralph
snwfdhmp/awesome-ralph
GitHub
html-meta
snwfdhmp/awesome-ralph
2026-09-04T05:23:38
ale-0980
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
Awesome Agent Loops
https://github.com/serenakeyitan/awesome-agent-loops
external
github.com
Curated collection of /loop, /goal, and /schedule commands for Claude Code and Codex sourced from practitioner posts, organized around trigger, condition, and skill structure.
Curated collection of /loop, /goal, and /schedule commands for Claude Code and Codex sourced from practitioner posts, organized around trigger, condition, and skill structure.
Curated collection of /loop, /goal, and /schedule commands for Claude Code and Codex sourced from practitioner posts, organized around trigger, condition, and skill structure.
The trigger or cadence is explicit, making the workflow recurring rather than one-off. Curated collection of /loop, /goal, and /schedule commands for Claude Code and Codex sourced from practitioner posts, organized around trigger, condition, and skill structure.
Use Awesome Agent Loops to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,754
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1754
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
objective;trigger
builder
cross-layer
adjacent
curated-index
C
ok
https://github.com/serenakeyitan/awesome-agent-loops
GitHub - serenakeyitan/awesome-agent-loops: A curated collection of the best /loop, /goal, and /schedule uses for Claude Code & Codex — real commands sourced from Twitter/X. The awesome-list of agent loops. · GitHub
A curated collection of the best /loop, /goal, and /schedule uses for Claude Code & Codex — real commands sourced from Twitter/X. The awesome-list of agent loops. - serenakeyitan/awesome-agent-loops
serenakeyitan/awesome-agent-loops
GitHub
html-meta
serenakeyitan/awesome-agent-loops
2026-09-04T05:23:38
ale-0981
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
Awesome Loop Models
https://github.com/huskydoge/Awesome-Loop-Models
external
github.com
Dedicated catalog of architectures that reuse a learned layer, block, module, or operator within one forward process; use it for deeper model-level coverage while this repository focuses on the bridge to operational agent loops.
Dedicated catalog of architectures that reuse a learned layer, block, module, or operator within one forward process; use it for deeper model-level coverage while this repository focuses on the bridge to operational agent loops.
Dedicated catalog of architectures that reuse a learned layer, block, module, or operator within one forward process; use it for deeper model-level coverage while this repository focuses on the bridge to operational agent loops.
Connects neighboring ecosystems while preserving Loop Engineering as a narrower operating concept. Dedicated catalog of architectures that reuse a learned layer, block, module, or operator within one forward process; use it for deeper model-level coverage while this repository focuses on the bridge to operational agent...
Use Awesome Loop Models to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,755
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1755
2026-07-18
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder
model
adjacent
curated-index
C
ok
https://github.com/huskydoge/Awesome-Loop-Models
GitHub - huskydoge/Awesome-Loop-Models: A curated list of papers and selected technical blogs on Loop Models. · GitHub
A curated list of papers and selected technical blogs on Loop Models. - huskydoge/Awesome-Loop-Models
huskydoge/Awesome-Loop-Models
GitHub
html-meta
huskydoge/Awesome-Loop-Models
2026-09-04T05:23:38
ale-0982
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
harness-engineering (Ryan Lopopolo)
https://github.com/lopopolo/harness-engineering
external
github.com
Ryan Lopopolo's anthology, field guide, and agent context bundle for harness engineering, collecting primary sources on the layer directly beneath loop engineering.
Ryan Lopopolo's anthology, field guide, and agent context bundle for harness engineering, collecting primary sources on the layer directly beneath loop engineering.
Ryan Lopopolo's anthology, field guide, and agent context bundle for harness engineering, collecting primary sources on the layer directly beneath loop engineering.
Context is managed as durable loop state rather than a single prompt payload. Ryan Lopopolo's anthology, field guide, and agent context bundle for harness engineering, collecting primary sources on the layer directly beneath loop engineering.
Use harness-engineering (Ryan Lopopolo) to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,756
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1756
2026-07-22
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
context
builder
cross-layer
adjacent
curated-index
C
ok
https://github.com/lopopolo/harness-engineering
GitHub - lopopolo/harness-engineering: 🐎 Ryan Lopopolo’s anthology, field guide, and agent context bundle for harness engineering · GitHub
🐎 Ryan Lopopolo’s anthology, field guide, and agent context bundle for harness engineering - lopopolo/harness-engineering
lopopolo/harness-engineering
GitHub
html-meta
lopopolo/harness-engineering
2026-09-04T05:23:38
ale-0983
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
A Coding-Agent Reading List: Behind the Loops
https://insights.ml4trading.io/p/a-coding-agent-reading-list-behind
external
insights.ml4trading.io
Stefan Jansen's curated reading path of 60+ resources on coding-agent loops, organized to separate practitioner discourse, control-theory foundations, agent primitives, harness papers, adoption studies, and safety work by evidence type.
Stefan Jansen's curated reading path of 60+ resources on coding-agent loops, organized to separate practitioner discourse, control-theory foundations, agent primitives, harness papers, adoption studies, and safety work by evidence type.
Stefan Jansen's curated reading path of 60+ resources on coding-agent loops, organized to separate practitioner discourse, control-theory foundations, agent primitives, harness papers, adoption studies, and safety work by evidence type.
Connects neighboring ecosystems while preserving Loop Engineering as a narrower operating concept. Stefan Jansen's curated reading path of 60+ resources on coding-agent loops, organized to separate practitioner discourse, control-theory foundations, agent primitives, harness papers, adoption studies, and safety work by...
Use A Coding-Agent Reading List: Behind the Loops to reuse a concrete artifact or connect it to the wider ecosystem.
Contextual source from insights.ml4trading.io; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,757
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1757
2026-07-23
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder
cross-layer
adjacent
curated-index
C
ok
https://insights.ml4trading.io/p/a-coding-agent-reading-list-behind
A Coding-Agent Reading List: Behind the Loops
Loop engineering is only the surface. A reading path through the older control problems underneath — and the line between what a coding agent may change and the experimental decisions that determine whether a result is valid.
Stefan Jansen
insights.ml4trading.io
html-meta
2026-09-04T05:23:38
ale-0984
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
awesome-agent-harness
https://github.com/Picrew/awesome-agent-harness
external
github.com
Companion index scoped to the harness layer itself, collecting projects, tools, benchmarks, and practical guides rather than the operational loop built around them.
Companion index scoped to the harness layer itself, collecting projects, tools, benchmarks, and practical guides rather than the operational loop built around them.
Companion index scoped to the harness layer itself, collecting projects, tools, benchmarks, and practical guides rather than the operational loop built around them.
The work turns loop quality into a measurable task or score. Companion index scoped to the harness layer itself, collecting projects, tools, benchmarks, and practical guides rather than the operational loop built around them.
Use awesome-agent-harness to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,758
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1758
2026-09-01
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
workspace;verification
builder
cross-layer
adjacent
curated-index
C
ok
https://github.com/Picrew/awesome-agent-harness
GitHub - Picrew/awesome-agent-harness: An awesome list of Agent Harness engineering resources, including GitHub projects, tools, benchmarks, and practical guides. · GitHub
An awesome list of Agent Harness engineering resources, including GitHub projects, tools, benchmarks, and practical guides. - Picrew/awesome-agent-harness
Picrew/awesome-agent-harness
GitHub
html-meta
Picrew/awesome-agent-harness
2026-09-04T05:23:38
ale-0985
Adjacent Awesome Lists
adjacent-awesome-lists
List
🧭
Awesome-Memory-for-Agents
https://github.com/TsinghuaC3I/Awesome-Memory-for-Agents
external
github.com
Paper collection scoped to memory for language agents, useful as a companion index when the memory layer specifically is the part being designed.
Paper collection scoped to memory for language agents, useful as a companion index when the memory layer specifically is the part being designed.
Paper collection scoped to memory for language agents, useful as a companion index when the memory layer specifically is the part being designed.
Persistent memory is treated as an external runtime artifact. Paper collection scoped to memory for language agents, useful as a companion index when the memory layer specifically is the part being designed.
Use Awesome-Memory-for-Agents to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,759
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1759
2026-09-04
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
context
builder
cross-layer
adjacent
curated-index
C
ok
https://github.com/TsinghuaC3I/Awesome-Memory-for-Agents
GitHub - TsinghuaC3I/Awesome-Memory-for-Agents: A Collection of Papers about Memory for Language Agents · GitHub
A Collection of Papers about Memory for Language Agents - TsinghuaC3I/Awesome-Memory-for-Agents
TsinghuaC3I/Awesome-Memory-for-Agents
GitHub
html-meta
TsinghuaC3I/Awesome-Memory-for-Agents
2026-09-04T05:23:38
ale-0986
Explore And Reuse
explore-and-reuse
Template
🧾
Resource Atlas
https://chaoyue0307.github.io/awesome-loop-engineering/
external
chaoyue0307.github.io
Filter 995 resources by goal, loop layer, lifecycle stage, artifact type, evidence class, and search query.
Filter 995 resources by goal, loop layer, lifecycle stage, artifact type, evidence class, and search query.
Filter 995 resources by goal, loop layer, lifecycle stage, artifact type, evidence class, and search query.
Turns the evidence into an interactive atlas and structured data. Filter 995 resources by goal, loop layer, lifecycle stage, artifact type, evidence class, and search query.
Use Resource Atlas to reuse a concrete artifact or connect it to the wider ecosystem.
Reusable template, schema, checklist, or guide; signal comes from concrete adaptation and validation.
medium
README.md
1,767
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1767
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
objective
builder
cross-layer
enabling
reusable-artifact
A
ok
https://chaoyue0307.github.io/awesome-loop-engineering/
Awesome Loop Engineering
Explore 989 resources from model recurrence to governed agent operations, then build with 22 patterns, 22 contracts, and 8 runtime starters.
Chaoyue He
chaoyue0307.github.io
html-meta
2026-09-04T05:23:38
ale-0987
Explore And Reuse
explore-and-reuse
List
🧭
Hugging Face dataset
https://huggingface.co/datasets/cy0307/awesome-loop-engineering
external
huggingface.co
Query the full collection as generated CSV and JSONL tables with publication, evidence, and lifecycle fields.
Query the full collection as generated CSV and JSONL tables with publication, evidence, and lifecycle fields.
Query the full collection as generated CSV and JSONL tables with publication, evidence, and lifecycle fields.
Packages the evidence as queryable CSV and JSONL rather than only a rendered page. Query the full collection as generated CSV and JSONL tables with publication, evidence, and lifecycle fields.
Use Hugging Face dataset to reuse a concrete artifact or connect it to the wider ecosystem.
Contextual source from huggingface.co; useful for practice signals or boundary conditions, not independent validation.
contextual
README.md
1,768
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1768
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder
cross-layer
enabling
curated-index
C
ok
https://huggingface.co/datasets/cy0307/awesome-loop-engineering
cy0307/awesome-loop-engineering · Datasets at Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
Hugging Face
domain-fallback
2026-09-04T05:23:38
ale-0988
Explore And Reuse
explore-and-reuse
Template
🧾
Dataset export guide
data/README.md
local_path
Load, query, regenerate, and audit the CSV, JSONL, and Resource Atlas data.
Load, query, regenerate, and audit the CSV, JSONL, and Resource Atlas data.
Load, query, regenerate, and audit the CSV, JSONL, and Resource Atlas data.
Packages the evidence as queryable CSV and JSONL rather than only a rendered page. Load, query, regenerate, and audit the CSV, JSONL, and Resource Atlas data.
Use Dataset export guide to reuse a concrete artifact or connect it to the wider ecosystem.
Repository file; inspect the linked schema, example, guide, or implementation.
medium
README.md
1,769
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1769
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder
cross-layer
enabling
repository-native
A
local_ok
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/data/README.md
Dataset export guide
2026
GitHub
GitHub
repository
2026-09-04T05:23:38
ale-0989
Explore And Reuse
explore-and-reuse
Template
🧾
Runtime selection guide
meta/RUNTIME_SELECTION.md
local_path
Compare session, scheduled, CI, cron, and durable runtimes by persistence, isolation, permissions, and state.
Compare session, scheduled, CI, cron, and durable runtimes by persistence, isolation, permissions, and state.
Compare session, scheduled, CI, cron, and durable runtimes by persistence, isolation, permissions, and state.
Durable execution and replay are treated as first-class loop infrastructure. Compare session, scheduled, CI, cron, and durable runtimes by persistence, isolation, permissions, and state.
Use Runtime selection guide to reuse a concrete artifact or connect it to the wider ecosystem.
Repository file; inspect the linked schema, example, guide, or implementation.
medium
README.md
1,770
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1770
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
trigger;workspace;state
builder
cross-layer
enabling
repository-native
A
local_ok
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/meta/RUNTIME_SELECTION.md
Runtime selection guide
2026
GitHub
GitHub
repository
2026-09-04T05:23:38
ale-0990
Explore And Reuse
explore-and-reuse
Template
🧾
Future Directions agenda
FUTURE-DIRECTIONS.md
local_path
Turn 15 open problems into measurable studies, runtime projects, product pilots, and shared standards.
Turn 15 open problems into measurable studies, runtime projects, product pilots, and shared standards.
Turn 15 open problems into measurable studies, runtime projects, product pilots, and shared standards.
Turns open gaps into measurable research, infrastructure, and product directions. Turn 15 open problems into measurable studies, runtime projects, product pilots, and shared standards.
Use Future Directions agenda to reuse a concrete artifact or connect it to the wider ecosystem.
Repository file; inspect the linked schema, example, guide, or implementation.
medium
README.md
1,771
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1771
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder
cross-layer
enabling
repository-native
A
local_ok
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/FUTURE-DIRECTIONS.md
Future Directions agenda
2026
GitHub
GitHub
repository
2026-09-04T05:23:38
ale-0991
Shape What Comes Next
shape-what-comes-next
Template
🧾
Release notes
https://github.com/ChaoYue0307/awesome-loop-engineering/releases
external
github.com
Versioned changelog of new resources, patterns, and repository changes.
Versioned changelog of new resources, patterns, and repository changes.
Versioned changelog of new resources, patterns, and repository changes.
Turns open questions and operating lessons into visible next work. Versioned changelog of new resources, patterns, and repository changes.
Use Release notes to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,786
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1786
2026-07-15
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder
cross-layer
enabling
reusable-artifact
A
ok
https://github.com/ChaoYue0307/awesome-loop-engineering/releases
Releases · ChaoYue0307/awesome-loop-engineering · GitHub
🔁 Build reliable recurring AI-agent systems: 989 resources, 22 operational patterns, 22 loop contracts, 8 runtime starters, an interactive atlas, and a structured dataset. - Releases · ChaoYue0307/awesome-loop-engineering
GitHub Releases
GitHub
html-meta
ChaoYue0307/awesome-loop-engineering
2026-09-04T05:23:38
ale-0992
Shape What Comes Next
shape-what-comes-next
Template
🧾
Roadmap
ROADMAP.md
local_path
Near-term work, pattern priorities, gallery goals, and open questions.
Near-term work, pattern priorities, gallery goals, and open questions.
Near-term work, pattern priorities, gallery goals, and open questions.
Turns open questions and operating lessons into visible next work. Near-term work, pattern priorities, gallery goals, and open questions.
Use Roadmap to reuse a concrete artifact or connect it to the wider ecosystem.
Repository file; inspect the linked schema, example, guide, or implementation.
medium
README.md
1,787
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1787
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
objective
builder
cross-layer
enabling
repository-native
A
local_ok
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/ROADMAP.md
Roadmap
2026
GitHub
GitHub
repository
2026-09-04T05:23:38
ale-0993
Shape What Comes Next
shape-what-comes-next
Template
🧾
Launch article
posts/launch.md
local_path
Concise explanation of the concept, implementation kit, and evidence base.
Concise explanation of the concept, implementation kit, and evidence base.
Concise explanation of the concept, implementation kit, and evidence base.
Turns open questions and operating lessons into visible next work. Concise explanation of the concept, implementation kit, and evidence base.
Use Launch article to reuse a concrete artifact or connect it to the wider ecosystem.
Repository file; inspect the linked schema, example, guide, or implementation.
medium
README.md
1,788
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1788
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder
cross-layer
enabling
repository-native
A
local_ok
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/posts/launch.md
Launch article
2026
GitHub
GitHub
repository
2026-09-04T05:23:38
ale-0994
Shape What Comes Next
shape-what-comes-next
Template
🧾
Discussion guide
meta/DISCUSSIONS.md
local_path
Suggested discussion categories, starter prompts, and moderation standard.
Suggested discussion categories, starter prompts, and moderation standard.
Suggested discussion categories, starter prompts, and moderation standard.
The resource is directly reusable as a starting artifact. Suggested discussion categories, starter prompts, and moderation standard.
Use Discussion guide to reuse a concrete artifact or connect it to the wider ecosystem.
Repository file; inspect the linked schema, example, guide, or implementation.
medium
README.md
1,789
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1789
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder
cross-layer
enabling
repository-native
A
local_ok
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/meta/DISCUSSIONS.md
Discussion guide
2026
GitHub
GitHub
repository
2026-09-04T05:23:38
ale-0995
Shape What Comes Next
shape-what-comes-next
Pattern
🔁
Show your Loop Engineering patterns
https://github.com/ChaoYue0307/awesome-loop-engineering/discussions/2
external
github.com
Community discussion for real or anonymized loop examples.
Community discussion for real or anonymized loop examples.
Community discussion for real or anonymized loop examples.
Turns open questions and operating lessons into visible next work. Community discussion for real or anonymized loop examples.
Use Show your Loop Engineering patterns to reuse a concrete artifact or connect it to the wider ecosystem.
Inspectable GitHub source; popularity is context, not proof of reliability.
medium
README.md
1,790
https://github.com/ChaoYue0307/awesome-loop-engineering/blob/main/README.md#L1790
Apply
apply
Reuse, adapt, and contribute concrete loop artifacts.
whole-loop
builder
cross-layer
enabling
operational-pattern
B
ok
https://github.com/ChaoYue0307/awesome-loop-engineering/discussions/2
Show your Loop Engineering patterns · ChaoYue0307/awesome-loop-engineering · Discussion #2 · GitHub
Show your Loop Engineering patterns
GitHub Discussions
GitHub
html-meta
ChaoYue0307/awesome-loop-engineering
2026-09-04T05:23:38