Cockroach Labs' five-month hospital-model workflow — treating bugs as patients with coding agents as a triage/treatment team — gets replicated or adopted by other engineering orgs as a working agentic bug-management practice.
state: watchingheat: mediumuncertainty: mediumconvergesscott: highagent-workflows coding-agents engineering-practiceCockroach Labs
What is this?
Cockroach Labs (the distributed SQL database company) published a first-party engineering report describing a 5-month production experiment: a coding-agent pipeline organized around a teaching-hospital metaphor — Triage Nurses, Fellows, Review Attendings, Discharge Nurses — that enforces mandatory planning, rigorous reviews, and safety guardrails before merge. They claim the pipeline processed >1M lines of code with only 7 reverts. The case hypothesis is whether this structured, role-based agent workflow gets replicated or adopted by other engineering organizations as a working agentic bug-management practice. The web evidence is the Cockroach Labs blog post (dated ~23 hours before search) plus a Register article and a LinkedIn post by the engineer (Rafi Shamim) describing 'MOLT Sinai'; no independent replication evidence appears in the snippets.
Why it matters to Scott
Cockroach Labs — a serious distributed-SQL infrastructure company — independently implemented a production agent workflow that embodies multiple load-bearing Scott frameworks: Micro-Agents Architecture (named roles with explicit handoffs), Long-Running Agents (stateless workers + durable external state, handover notes), Scout–Senior Split (commodity triage → frontier review), Agent Loop (PLAN→EVALUATE→GOVERN gates), and Evaluation-Driven Development / Verification Loops (mandatory planning, review gates, 7 reverts/1M LOC). The hospital metaphor (Triage Nurses, Fellows, Review Attendings, Discharge Nurses) is a concrete instantiation of Role-gated responsibility workflow and Handover-Note Discipline. The case hypothesis — whether this replicates elsewhere — directly engages Scott's Exemplar Ratchet, Forward-Deployed Practice OS, and Learning Subsidy frameworks, which define what makes a team-internal workflow transferable. This is a dated-receipts moment: a consequential other party has arrived at Scott's architecture in production at scale.
ip:framework.micro-agents-architectureip:framework.long-running-agentsip:framework.scout-senior-splitip:framework.agent-loopip:framework.exemplar-ratchetip:framework.forward-deployed-practice-operating-systemip:concept.evaluation-driven-developmentip:concept.verification-loopsip:concept.handover-note-disciplineip:concept.plan-mode-disciplineip:concept.observable-autonomyip:concept.role-gated-responsibility-workflowradar:37signals-agent-driven-defaultradar:checkly-agentic-go-rewriteradar:chief-of-staff-agent-team-patternradar:bdfl-supervisor-workflow-validationradar:devcake-ticket-driven-software-factoryradar:agentlane-git-native-coordinationradar:keel-modular-agent-conductorradar:trueforge-open-agent-harnessradar:zuse-parallel-coding-workspacesradar:agent-handoff-protocol-adoption
queries asked of Scott's wikis
- agent-workflow patterns: multi-agent orchestration with named roles and handoffs
- coding-agent safety: mandatory planning, review gates, revert rates as eval metrics
- agentic engineering practice adoption: what makes a team-internal workflow replicable elsewhere
- agent memory / context passing across role boundaries in a pipeline
- local vs. cloud inference economics for sustained high-volume agent pipelines
- Scott's own agent-harness evaluations: criteria for 'this works, ship it'
Measured heat
now 8 pts/hpeak 22 pts/hcomments 4/hpeers p87momentum: steady1 platformsage 49h
points/hour across evidence · reading as of 2026-10-12 02:59:37.977291+11:00 · deterministic, not a model opinion
How the heat travelled
pace: p80 vs 1204 stories at the 48h mark (now 49h old) — ahead of gemini-38-flash-coding-gap (1.0x), behind gitspawn-repository-agent-hijack (1.0x)
Evidence (1) — ⭐ canonical anchor
Interpretation history
2026-10-10T00:08:13Z
grounded: converges/high — Cockroach Labs — a serious distributed-SQL infrastructure company — independently implemented a production agent workflow that embodies multiple load-bearing Sc
2026-10-09T23:58:22Z
case created — First-party engineering report of sustained, structured agent workflow at serious infrastructure company; adoption evidence Scott tracks.
Decision trace
- 10-12 02:35sensor_dirtycomment_update
- 10-11 22:31sensor_dirtycomment_update
- 10-11 21:33sensor_dirtyvelocity_spike
- 10-11 15:03feedback_briefingScott vote via UI
- 10-11 14:27sensor_dirtycomment_update
- 10-11 13:31sensor_dirtyvelocity_spike
- 10-11 10:29sensor_dirtycomment_update
- 10-10 18:57attention_communicatedCockroach Labs published a first-party account of a sustained agent workflow where bugs are 'patients' and coding agents staff a medical team: Triage Nurses classify, Fellows implement, Revi
- 10-10 18:57attention_routeNew since the last package and the strongest independent validation yet of Scott's named-role and handoff architecture at scale. It belongs in the briefing as a lead — nothing here has an action
- 10-10 18:04attention_routeConsequential independent validation of multiple Scott frameworks in production at a respected infrastructure company. Belongs in the briefing as a lead story for Scott's Exemplar Ratchet and For
- 10-10 11:23attention_routeConsequential independent validation of multiple Scott frameworks in production at a respected infrastructure company. Belongs in the next briefing as a lead story for Scott's Exemplar Ratchet an
- 10-10 11:18attention_candidatecreate
- 10-10 11:08groundCockroach Labs — a serious distributed-SQL infrastructure company — independently implemented a production agent workflow that embodies multiple load-bearing Scott frameworks: Micro-Agents Architectur
- 10-10 10:58createFirst-party engineering report of sustained, structured agent workflow at serious infrastructure company; adoption evidence Scott tracks.