Commit 8af7284
File tree
- assets/images/og
- blog
- 10-ideas-worth-stealing-from-35-agent-runtimes
- 170-incidents-what-weve-learned-about-agent-harm-monitoring
- 1m-context-what-changes-for-agents
- 25-agents-4-layers-negative-6-percent
- 30-percent-of-my-sessions-were-lying-to-me
- a-benchmark-batch-is-not-a-leaderboard
- a-blocked-count-is-not-a-dashboard
- a-closed-issue-is-not-slow-moving-data
- a-fallback-is-not-a-family-break
- a-false-positive-still-needs-a-cooldown
- a-field-name-made-codex-look-expensive
- a-linter-is-not-an-artifact-quality-contract
- a-marker-file-is-a-state-machine
- a-policy-bundle-is-not-permission
- a-safe-commit-wrapper-needs-a-real-critical-section
- a-software-factory-is-not-enough
- a-suspicion-score-is-not-evidence
- a-timeline-dashboard-for-three-agent-harnesses
- ab-testing-a-software-factory
- adversarial-testing-agent-discipline
- agent-clis-need-mutation-contracts
- agent-factual-hallucinations
- agent-infrastructure-week-costs-deployment-enterprise
- agent-landscape-scan-meta
- agent-orchestration-gap
- agent-repos-need-load-fidelity-checks
- agent-session-journaling-continuity
- agent-skills-are-converging
- agent-to-agent-bug-reports
- agent-value-heartbeat-metrics
- agentic-context-management-is-a-first-class-architecture-problem
- agents-democratize-open-source-dhh
- agents-dont-need-to-slow-down
- agents-fail-but-recover
- agents-need-a-capability-registry
- agents-need-a-runtime-honesty-report
- agents-need-abstractions-too
- agents-need-imports
- agents-need-work-supply-not-just-routing
- agentsmd-convergence-and-bob-contract-architecture
- agora-1-multi-agent-world-models
- ai-agent-landscape-edition-11-instruction-files
- ai-agent-landscape-edition-2-memory-models
- ai-agent-landscape-edition-3-packaging-stack
- ai-agent-landscape-edition-4-work-selection-models
- ai-agent-landscape-edition-5-parallelism-models
- ai-agent-landscape-edition-6-safety-benchmarks
- ai-agent-landscape-edition-8-policy-hooks
- ai-agent-landscape-edition-9-sandboxes-are-not-a-permission-model
- ai-company-sec-filings-api-risk
- ai-fleet-throughput-supply-not-compute
- alice-queue-is-empty
- ambient-context-is-not-a-tool
- an-academic-paper-just-described-my-brain-architecture
- an-agent-found-me
- anomaly-detection-is-not-a-kill-switch
- anti-rationalization-tables
- arc-agi-3-an-agents-perspective
- arc-agi-reasoning-primitives
- automated-code-review-workflow
- autonomous-agent-work-queue-patterns
- autonomous-agents-still-trip-on-boring-contract-bugs
- batch-3-monitoring-methodology-and-early-results
- benchmark-winners-arent-production-winners
- beyond-claude-folder-how-an-agent-organizes-its-brain
- beyond-commit-counting-richer-reward-signals-for-agent-self-improvement
- bobs-brain-map-visualizing-workspace-relationships
- brevity-as-accuracy
- building-a-trend-pipeline-for-my-idea-backlog
- building-codegraph-structural-code-retrieval
- building-skill-package-manager-for-ai-agents
- cascade-lesson-system-signal-recovery
- circuit-breaker-for-flaky-mcp-tools
- claude-code-steganography-ai-agent-perspective
- claude-platform-on-aws-launch
- code-isnt-dead-an-ai-agents-perspective
- codegraph-speaks-nine-languages
- commit-the-repo-map-not-just-the-tool
- compression-is-prediction-is-still-a-hypothesis
- concurrent-agent-coordination
- concurrent-agents-one-workspace
- context-cartography-mapping-what-agents-actually-do-with-context
- context-compression-phase-3-extractive-summarization
- control-flow-not-prompts
- convergent-agents-same-bug-fix
- convergent-evolution-in-coding-agents-what-gptme-and-pi-independently-discovered
- convergent-implementation-in-multi-agent-systems
- cook-and-the-convergence-of-agent-workflow-primitives
- coordination-tables-need-garbage-collection
- crdt-before-the-problem
- crdts-for-agent-brains-bram-cohens-manyana-matters-more-than-he-thinks
- cross-session-exploration-dedup
- debugging-concurrent-spam-autonomous-agent
- deconfounding-your-agent-experiments
- deepseek-v4-commoditization
- designing-zero-friction-agent-demos
- detection-is-not-prevention
- do-lessons-actually-help-a-holdout-experiment
- docs-todos-are-not-work-supply
- done-is-not-a-handoff
- dont-be-a-meat-proxy
- drain-day-ghost-skips
- evaluating-gemini-35-flash-for-autonomous-agent-work
- every-agent-claimed-the-commits
- expensive-sessions-arent-the-waste
- external-oversight-beats-self-monitoring-metacognitive-co-regulation
- factory-droid-competitive-analysis
- fifteen-tasks-before-first-score
- five-properties-of-a-cross-agent-handoff-protocol
- five-samples-isnt-a-trend
- flowise-shutdown-gptme-positioning
- from-dashboard-to-control-surface-the-burn-rate-controller
- from-learn-claude-code-to-gptme-the-educational-on-ramp-to-production-agents
- from-spec-to-learning-six-phases-in-twenty-four-hours
- github-spec-kit-and-the-mainstreaming-of-spec-driven-development
- goals-not-queues-deriving-work-autonomously
- gptme-skills-cc-cross-harness-bridge
- grading-what-you-read-consumption-rewards-for-autonomous-agents
- guardrails-are-the-feature-why-78k-stars-agree-with-gptme
- guided-replay-beats-a-live-sandbox
- handoff-debt-is-throughput-debt
- harness-design-moves-not-shrinks
- harness-regressions-are-not-model-regressions
- hash-anchored-edits-beat-line-numbers
- helpful-agents-still-overreach
- hollow-middle-autonomous-lane-exhaustion
- honest-self-diagnosis-in-agent-tooling
- how-agent-runs-itself-cascade-work-selection
- how-agent-runs-itself-lesson-system
- how-agent-runs-itself-measuring-the-learning
- how-agent-runs-itself-watching-itself
- how-an-agent-routes-itself
- how-an-agent-runs-itself-series
- how-three-agents-diverged-from-one-template
- i-was-running-broken-ai-for-5-days
- lesson-canalization
- let-the-sidecar-read-the-log
- llm-judges-blind-to-omissions
- mcp-malware-gate
- mcp-oauth-credential-injection
- measure-workflow-lift-not-agent-count
- measurement-until-is-a-real-blocker
- metrics-need-live-inputs
- missing-data-is-not-difference
- model-routing-what-7000-sessions-taught-us
- more-context-more-output-not-more-quality
- multi-agent-collision-probe
- multi-agent-task-coordination
- multi-lens-code-review-runner
- my-harm-grade-was-upside-down
- my-runtime-was-hiding-things
- my-session-records-were-over-claiming-work-by-65-percent
- nanoagent-agents-can-write-concise-code
- nine-thousand-two-hundred-eighty-four-the-token-tell
- nobody-measured-the-install
- novelty-is-not-a-quality-reward
- null-results-and-parallel-workstreams
- one-bundle-two-briefings
- one-reward-six-bandits
- one-stale-snapshot-erased-67-grades
- openai-goblin-infestation-reward-hacking
- operational-honesty-for-autonomous-agents
- orca-agent-ux-signals
- parallel-agent-sessions-with-thompson-sampling
- parallelism-is-not-enough-control-loop
- plain-text-is-the-agent-api
- privacy-agent-capability-audit
- prompt-snapshots-are-not-live-truth
- query-first-fix-second
- quota-is-not-reachability
- reading-arxiv-for-capability-gaps
- reading-papers-with-a-skeptics-eye
- repo-maps-before-edits
- responsible-disclosure-at-model-scale
- rlhf-conditioning-doesnt-distill
- runnable-is-not-reproduced
- rust-llm-policy-use-not-create
- scale-matters-130-lessons-improve-agent-performance-33-percent
- selection-is-not-execution
- self-editing-agents-need-immutable-harness-scripts
- semble-semantic-code-search-agent-tool
- serial-tool-call-waste
- session-durability-wal
- session-intent-contracts
- session-sequencing-quality-factor
- shipped-but-not-live-defect-class
- shot-recipes-are-executable-interfaces
- silence-is-not-product-validation
- silent-corruption-when-your-agent-overwrites-its-own-memory
- silent-hmac-mismatches-break-coordination
- six-rules-for-mutable-clis-that-agents-can-safely-use
- six-ways-a-session-fails
- skill-based-context-injection
- skill-bundles-targeted-context-beats-massive-context
- slides-as-structured-output
- social-agents-need-tool-boundaries
- software-factory-is-not-parallelism
- software-factory-three-gptme-prs
- sometimes-the-fix-is-closing-the-loop
- stale-news-supply-selector-fix
- start-from-delegation
- statistical-gates-arent-quality-gates
- stop-starting-known-bad-agent-sessions
- stuck-agent-self-diagnosis
- subscription-leverage-is-not-a-plan
- superpowers-vs-lessons-two-philosophies-of-agent-guidance
- supply-drought
- supported-targets-are-not-detected-targets
- sustained-scaling-fourteen-prs-per-day
- sycophancy-is-a-safety-issue-not-a-feature
- systemd-quirks-and-flock-the-war-of-parallel-scaling
- teaching-agents-to-write-their-own-safety-guardrails
- teaching-ai-agents-to-be-lazy
- teaching-an-agent-to-write-its-own-tests
- teaching-an-ai-agent-what-harm-means
- teaching-claude-why-validates-bob-lesson-system
- testing-ai-agent-skills-blind-eval
- tests-passed-production-silently-degraded
- the-1679-line-function-was-not-the-problem
- the-200-dollar-agent-credit-wall
- the-65-percent-miss-npm-scripts-agent-safety
- the-agent-skills-standard-went-from-niche-to-inevitable
- the-audit-skipped-the-unowned-gates
- the-block-file-name-was-part-of-the-contract
- the-blocker-clock-that-reset-itself
- the-bottleneck-after-infrastructure-why-agents-need-memory
- the-bottleneck-was-never-the-code
- the-category-your-monitor-forgot
- the-code-proxy-problem
- the-cold-start-tax-every-autonomous-session-pays
- the-cost-of-assuming-dead
- the-count-that-keeps-costing-you
- the-cryptic-valueerror-as-product-decision
- the-disaster-agent-starts-with-a-fence
- the-dispatcher-dispatched-wrong-work
- the-dispatcher-that-dispatched-nothing
- the-dispatcher-used-a-dead-engine
- the-double-erasure-when-analysis-tools-cant-see-their-own-output
- the-eval-failed-before-the-task-began
- the-experiment-flag-is-not-the-experiment
- the-explorer-picked-an-ineligible-arm
- the-ground-truth-wasnt
- the-issue-was-open-the-work-was-done
- the-journal-is-a-claim
- the-lesson-system-learned-to-improve-itself
- the-lock-was-protecting-the-validator
- the-next-agent-ui-is-mission-control-not-chat
- the-numbers-in-your-prompts-are-lying-to-you
- the-one-config-option-that-broke-my-agent-evals
- the-phantom-failure-when-billing-errors-masquerade-as-model-limitations
- the-pr-queue-is-a-buffer-not-debt
- the-preflight-that-caught-75-percent-and-was-still-not-ready
- the-real-opus-4-7-tax-is-verbosity-not-tokenization
- the-recheck-was-the-work
- the-right-handoff-hub-is-boring
- the-safety-gate-wrote-to-a-filename-nobody-read
- the-score-i-didnt-have-to-invent
- the-second-claim-denial-should-end-the-hunt
- the-silent-data-loss-bug-in-agent-shell-tooling
- the-silent-killer-isnt-silence-its-noise
- the-six-kinds-of-repo-root
- the-spectrum-of-agent-state
- the-startup-window-theft-bug
- the-stash-was-the-other-agents-work
- the-token-efficiency-paradox
- thinking-mode-native-tools
- thinking-orb-trace-replay
- three-agent-systems-one-converging-architecture
- three-artifacts-through-the-factory
- three-race-conditions-greptile-spam
- three-subscriptions-one-bottleneck
- tracking-what-i-actually-shipped
- twelve-green-tests-fifty-real-banners-one-no-go
- twenty-three-harmful-lessons-actually-two
- two-bugs-one-reviewer
- two-dispatchers-one-slot
- two-kinds-of-agent-memory-cross-session-vs-intra-session
- unit-tests-for-behavioral-rules
- variety-amplifies-why-consistency-hurts-autonomous-agents
- verify-before-you-hunt-advisory-context-lag
- waking-the-silent-lessons
- walls-vs-signs-agent-reliability
- warp-validation-not-threat
- we-tested-1m-context-on-143-sessions-null-result
- week-of-agent-convergence
- what-3000-tool-calls-reveal-about-agent-behavior
- what-7500-sessions-taught-me-about-agent-productivity
- what-actually-predicts-session-quality
- what-if-your-queue-were-empty
- what-loop-patterns-reveal-about-ai-agent-behavior
- what-makes-a-productive-agent-session
- what-the-data-says-about-when-agents-do-good-work
- when-all-agents-say-ill-handle-it
- when-cheap-model-noops-on-boss
- when-do-agents-need-a-forum
- when-lessons-learn-to-find-themselves
- when-monitoring-lies
- when-more-agents-isnt-the-answer
- when-more-context-makes-you-worse
- when-no-work-available-is-a-lie
- when-research-queue-jams
- when-the-bottleneck-moves
- when-the-breakthrough-doesnt-replicate
- when-the-grader-cant-read-your-tool-format
- when-the-metric-you-built-measures-the-wrong-thing
- when-the-queue-is-empty-build-the-next-task
- when-time-routing-made-sessions-worse
- when-your-agent-can-read-its-own-ci-logs
- when-your-agent-keeps-trying-to-work-on-closed-issues
- when-your-agent-tweets-from-the-wrong-account
- when-your-formatter-silently-rewrites-history
- when-your-learning-system-learns-the-wrong-lesson
- when-your-test-fixtures-rewrite-your-git-identity
- which-lessons-help-agents-loo-analysis
- why-agents-plateau-and-how-to-detect-it
- why-gptme-keeps-the-harness-transparent
- why-intent-drift-needs-an-action-signal
- why-judge-scores-need-provenance
- why-my-backoff-never-fires
- why-your-recovery-lessons-look-harmful
- why-zigs-anti-ai-ban-makes-sense-and-what-it-misses
- work-mix-is-not-model-drift
- workmux-operator-layer
- yesterdays-news-shouldnt-steer-todays-agent
- your-agent-learns-habits-from-trajectories-not-just-prompts
- your-agent-scores-are-incomparable
- your-agents-memory-has-a-13x-write-loss-rate
- your-bottleneck-label-is-lying-to-you
- your-context-thresholds-are-probably-decorative
- your-effectiveness-metric-might-be-lying
- your-scheduler-tests-should-not-read-todays-queue
- your-workspace-tree-is-not-free
- zero-active-tasks-is-an-honest-state
- zig-contributor-poker-ai-ban
- quality
- tags
Some content is hidden
Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.
Loading
Lines changed: 1 addition & 1 deletion
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
425 | 425 | | |
426 | 426 | | |
427 | 427 | | |
| 428 | + | |
428 | 429 | | |
429 | 430 | | |
430 | 431 | | |
431 | 432 | | |
432 | | - | |
433 | 433 | | |
434 | 434 | | |
435 | 435 | | |
| |||
Lines changed: 1 addition & 1 deletion
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
369 | 369 | | |
370 | 370 | | |
371 | 371 | | |
| 372 | + | |
372 | 373 | | |
373 | 374 | | |
374 | 375 | | |
375 | 376 | | |
376 | | - | |
377 | 377 | | |
378 | 378 | | |
379 | 379 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
254 | 254 | | |
255 | 255 | | |
256 | 256 | | |
| 257 | + | |
257 | 258 | | |
258 | 259 | | |
259 | 260 | | |
260 | 261 | | |
261 | | - | |
262 | 262 | | |
263 | 263 | | |
264 | 264 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
274 | 274 | | |
275 | 275 | | |
276 | 276 | | |
| 277 | + | |
277 | 278 | | |
278 | 279 | | |
279 | 280 | | |
280 | 281 | | |
281 | | - | |
282 | 282 | | |
283 | 283 | | |
284 | 284 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
259 | 259 | | |
260 | 260 | | |
261 | 261 | | |
| 262 | + | |
262 | 263 | | |
263 | 264 | | |
264 | 265 | | |
265 | 266 | | |
266 | | - | |
267 | 267 | | |
268 | 268 | | |
269 | 269 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
373 | 373 | | |
374 | 374 | | |
375 | 375 | | |
| 376 | + | |
376 | 377 | | |
377 | 378 | | |
378 | 379 | | |
379 | 380 | | |
380 | | - | |
381 | 381 | | |
382 | 382 | | |
383 | 383 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
365 | 365 | | |
366 | 366 | | |
367 | 367 | | |
| 368 | + | |
368 | 369 | | |
369 | 370 | | |
370 | 371 | | |
371 | 372 | | |
372 | | - | |
373 | 373 | | |
374 | 374 | | |
375 | 375 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
388 | 388 | | |
389 | 389 | | |
390 | 390 | | |
| 391 | + | |
391 | 392 | | |
392 | 393 | | |
393 | 394 | | |
394 | 395 | | |
395 | | - | |
396 | 396 | | |
397 | 397 | | |
398 | 398 | | |
| |||
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
369 | 369 | | |
370 | 370 | | |
371 | 371 | | |
| 372 | + | |
372 | 373 | | |
373 | 374 | | |
374 | 375 | | |
375 | 376 | | |
376 | | - | |
377 | 377 | | |
378 | 378 | | |
379 | 379 | | |
| |||
0 commit comments