Independent use will determine whether OpenCode-memory reliably preserves and retrieves useful coding-agent context across OpenCode sessions while remaining practical to run locally.
state: expiredheat: lowuncertainty: highknownscott: lowagent-memory opencode coding-agentscioffiAIOpenCode
What is this?
OpenCode-memory is presented as a local-first persistent-memory layer for the OpenCode coding agent, intended to capture technical context when sessions go idle and retrieve or inject it in later sessions. The supplied snippets describe directory-based project identity and local retrieval approaches, including vector search and hybrid dense/BM25 indexing, but appear to cover several similarly named OpenCode memory projects. The case attributes the initial release to Antonio Cioffi/cioffiAI, while the strongest GitHub snippet points to tickernelz/opencode-mem, so authorship, exact implementation details, and real-world reliability are not firmly established by this material; independent use would be needed to validate them.
Why it matters to Scott
Scott already holds the core position in “Long-Running Agents” and “Session Isolation”: coding-agent continuity requires durable external state that a fresh session can selectively reload. OpenCode-memory is currently another unvalidated implementation of that pattern, closely duplicating the radar’s existing Mnemara evaluation case; without independent reliability or retrieval results, it does not yet alter his active memory-system designs or claims.
ip:framework.long-running-agentsip:concept.session-isolationip:framework.context-engineeringdev:project.search-conversationsradar:concept.agent-memoryradar:mnemara-claude-agent-memoryradar:self-written-notes-reasoning-gains
queries asked of Scott's wikis
- coding-agent persistent memory across sessions
- agent-maintained project memory and wikis
- automatic memory capture versus explicit curation
- local-first RAG for coding agents
- memory retrieval quality and context-injection costs
- directory-scoped identity for agent context
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
2026-08-13T09:35:03Z
No independent use or implementation evidence emerged within the observation window, and the release drew no further discussion. The artifact remains unvalidated, but no longer warrants an active episode absent future reliability results.
2026-08-11T08:31:33Z
No independent use, implementation results, or new discussion has appeared; the case remains an unvalidated local-memory release duplicating an already tracked pattern.
2026-08-11T08:29:38Z
grounded: known/low — Scott already holds the core position in “Long-Running Agents” and “Session Isolation”: coding-agent continuity requires durable external state that a fresh ses
2026-08-11T08:26:18Z
origin walked (codex/luna, conf 0.97): anchor hn.story.49254884 -> echo.github.7b18658e88 by Antonio Cioffi
2026-08-11T08:25:11Z
case created — A concrete open-source artifact creates a distinct validation episode, but it currently has only one low-engagement observation.
Decision trace
- 08-13 19:35expireNo independent use or implementation evidence emerged within the observation window, and the release drew no further discussion. The artifact remains unvalidated, but no longer warrants an active epis
- 08-13 19:35alert_silentThe staleness trigger adds no consequential evidence; Scott can ignore this implementation unless independent cross-session retrieval results appear later.
- 08-13 19:35alert_routeThe staleness trigger adds no consequential evidence; Scott can ignore this implementation unless independent cross-session retrieval results appear later.
- 08-11 18:31repriceNo independent use, implementation results, or new discussion has appeared; the case remains an unvalidated local-memory release duplicating an already tracked pattern.
- 08-11 18:31alert_silentThe reobservation is unchanged and adds no evidence about cross-session retrieval reliability or practical local operation, so the normal briefing is sufficient.
- 08-11 18:31alert_routeThe reobservation is unchanged and adds no evidence about cross-session retrieval reliability or practical local operation, so the normal briefing is sufficient.
- 08-11 18:30alert_silentThe public release is established, but it is another unvalidated local persistent-memory implementation of a pattern Scott already tracks. No independent evidence yet shows reliable cross-session retr
- 08-11 18:30alert_routeThe public release is established, but it is another unvalidated local persistent-memory implementation of a pattern Scott already tracks. No independent evidence yet shows reliable cross-session retr
- 08-11 18:29groundScott already holds the core position in “Long-Running Agents” and “Session Isolation”: coding-agent continuity requires durable external state that a fresh session can selectively reload. OpenCode-me
- 08-11 18:26promote_anchororigin walk conf 0.97
- 08-11 18:25createA concrete open-source artifact creates a distinct validation episode, but it currently has only one low-engagement observation.