InfiniteMemOs maintainer Marco Tessari claims the released deterministic episodic-memory system scores 70.49 on LoCoMo, potentially providing agents with more reproducible long-term recall than conventional probabilistic memory pipelines.
state: expiredheat: lowuncertainty: highknownscott: lowagent-memory coding-agentsMarco TessariInfiniteMemOs
What is this?
InfiniteMemOs is presented as a released episodic-memory system maintained by Marco Tessari, with a claimed LoCoMo score of 70.49 and an emphasis on deterministic, reproducible recall for AI agents. The supplied search summary states those claims, while other snippets establish episodic memory as structured storage and retrieval of past agent experiences and show competing systems reporting higher LoCoMo scores. However, none of the supplied result snippets directly verifies the repository, maintainer, architecture, evaluation setup, or whether its 70.49 result is comparable to other LoCoMo claims.
Why it matters to Scott
Scott already holds the deterministic-components-for-reliability position in `ip:concept.deterministic-ai-pendulum`, and the radar already tracks the near-identical deterministic-first memory claim in `radar:cuemap-deterministic-agent-memory`. InfiniteMemOs is therefore another example of an established pattern; without verified architecture, provenance behavior, or a comparable LoCoMo evaluation, it does not yet bear materially on Scott’s dev-wiki or memory designs.
ip:concept.deterministic-ai-pendulumdev:project.dev-wikiradar:cuemap-deterministic-agent-memoryradar:agent-memory-leaderboard-validationradar:concept.agent-memory
queries asked of Scott's wikis
- deterministic vs probabilistic agent memory
- episodic memory for coding agents
- reproducible retrieval and memory pipelines
- LoCoMo benchmark validity and comparability
- agent-maintained wikis as long-term memory
- memory provenance and deterministic recall
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
2026-09-05T22:23:07Z
At the 48-hour stale horizon, this remains a maintainer claim echoed across two records, not independent corroboration of a release or reproducible memory advantage. With no substantive follow-up or expected confirming event, retire active tracking; architecture details or a reproducible evaluation could reopen it.
2026-09-03T21:36:57Z
No new evidence or discussion has emerged; this remains an unvalidated benchmark claim and another example of an already tracked deterministic-memory pattern. Promotion now depends on reproducible evaluation details, independent testing, or architecture showing a practical advantage.
2026-09-03T21:31:24Z
grounded: known/low — Scott already holds the deterministic-components-for-reliability position in `ip:concept.deterministic-ai-pendulum`, and the radar already tracks the near-ident
2026-09-03T21:27:51Z
case created — The released implementation and numerical benchmark claim form a bounded agent-memory episode suitable for reproduction and comparison.
Decision trace
- 09-06 08:23expireAt the 48-hour stale horizon, this remains a maintainer claim echoed across two records, not independent corroboration of a release or reproducible memory advantage. With no substantive follow-up or e
- 09-06 08:23alert_silentThere is no new technical or access delta to surface, and the original testimony does not establish a consequential change for Scott’s memory designs. No specific confirmation is expected within six h
- 09-06 08:23alert_routeThere is no new technical or access delta to surface, and the original testimony does not establish a consequential change for Scott’s memory designs. No specific confirmation is expected within six h
- 09-04 07:36repriceNo new evidence or discussion has emerged; this remains an unvalidated benchmark claim and another example of an already tracked deterministic-memory pattern. Promotion now depends on reproducible eva
- 09-04 07:36alert_silentThe only delta is an unchanged reobservation, with no technical validation, adoption, or independent corroboration; nothing new warrants interrupting Scott or holding for an imminent confirmation.
- 09-04 07:36alert_routeThe only delta is an unchanged reobservation, with no technical validation, adoption, or independent corroboration; nothing new warrants interrupting Scott or holding for an imminent confirmation.
- 09-04 07:31alert_silentThe repository establishes that InfiniteMemOs was released and that its maintainer reports a 70.49 LoCoMo score, but the capability claim is unvalidated and adds no material engineering delta beyond a
- 09-04 07:31alert_routeThe repository establishes that InfiniteMemOs was released and that its maintainer reports a 70.49 LoCoMo score, but the capability claim is unvalidated and adds no material engineering delta beyond a
- 09-04 07:31groundScott already holds the deterministic-components-for-reliability position in `ip:concept.deterministic-ai-pendulum`, and the radar already tracks the near-identical deterministic-first memory claim in
- 09-04 07:27createThe released implementation and numerical benchmark claim form a bounded agent-memory episode suitable for reproduction and comparison.