Independent use will determine whether oh-my-subagents can reliably execute multi-day, subagent-driven codebase refactors with manageable human supervision.
state: expiredheat: lowuncertainty: highknownscott: lowagent-harnesses coding-agents long-horizon-agentsringlochid
What is this?
oh-my-subagents is presented in a Show HN post by ringlochid as a subagent workflow that refactored a codebase over three days. The supplied search snippets describe the broader orchestrator/worker pattern—scoped task decomposition, isolated work, verification loops, and human approval—but provide no independent evidence about this specific tool’s reliability, supervision cost, or results. The “AutoClaw prototype snapshot” evidence title is too truncated to establish its relationship to oh-my-subagents, so the case remains an author-reported experiment awaiting independent replication.
Why it matters to Scott
Scott already holds the relevant position in Long-Running Agents, Evaluation-Driven Development, and Discussed Is Not Deployed: multi-day agent claims require durable state, explicit completion evidence, and repeatable independent evaluation. This author-reported experiment is closely analogous to existing radar cases on supervised repository-improvement harnesses, but without independent results it adds only another unverified example.
ip:framework.long-running-agentsip:concept.evaluation-driven-developmentip:framework.discussed-is-not-deployedip:framework.micro-agents-architectureradar:e3d-pilot-sha-gated-agent-harnessradar:flow-claude-code-supervisorradar:concept.long-horizon-agentsradar:concept.agent-evaluation
queries asked of Scott's wikis
- long-horizon coding-agent harnesses
- subagent task decomposition and isolation
- agent supervision cost and approval boundaries
- verification loops for autonomous refactors
- persistent context for multi-day agents
- independent evaluation of coding-agent claims
Measured heat
no measured readings yet — the hourly heat pass fills this in
How the heat travelled
no chain yet — the hourly chain pass fills this in
Evidence (2) — ⭐ canonical anchor
Interpretation history
2026-08-22T03:26:45Z
No independent use or validation emerged after more than a month, so the author-reported experiment has not developed into a broader long-horizon agent signal. The reliability hypothesis remains untested rather than disproved.
2026-08-22T03:26:05Z
grounded: known/low — Scott already holds the relevant position in Long-Running Agents, Evaluation-Driven Development, and Discussed Is Not Deployed: multi-day agent claims require d
2026-08-22T03:24:35Z
origin walked (codex/luna, conf 0.93): anchor hn.story.49396071 -> echo.github.d6608d6e69 by Leo Zhang (ringlochid)
2026-08-22T03:22:44Z
case created — A runnable first-party repository and reported three-day refactor establish a distinct but currently low-traction long-horizon coding-agent episode.
Decision trace
- 08-22 13:26expireNo independent use or validation emerged after more than a month, so the author-reported experiment has not developed into a broader long-horizon agent signal. The reliability hypothesis remains untes
- 08-22 13:26alert_silentThere is no new consequential delta: engagement is unchanged and no independent implementation, evaluation, or result has appeared.
- 08-22 13:26alert_routeThere is no new consequential delta: engagement is unchanged and no independent implementation, evaluation, or result has appeared.
- 08-22 13:26alert_silentA new repository and author-reported three-day refactor are established, but there are no independent results, concrete before/after evidence, or transferable implementation findings that materially a
- 08-22 13:26alert_routeA new repository and author-reported three-day refactor are established, but there are no independent results, concrete before/after evidence, or transferable implementation findings that materially a
- 08-22 13:26groundScott already holds the relevant position in Long-Running Agents, Evaluation-Driven Development, and Discussed Is Not Deployed: multi-day agent claims require durable state, explicit completion eviden
- 08-22 13:24promote_anchororigin walk conf 0.93
- 08-22 13:22createA runnable first-party repository and reported three-day refactor establish a distinct but currently low-traction long-horizon coding-agent episode.