Conversation
Surfaces two ranking-signal gaps from Zaid's 2026-04-28 article "Why Karpathy's Second Brain Breaks at Agent Scale": - #57: Provenance strength (stated vs inferred) is recorded but never factors into RRF ranking. Cheap win — data already exists. - #58: last_recalled_at only flags stale facts; reinforcement frequency and age decay don't influence retrieval order. Both medium priority; sit behind the 1.0 punchlist but ahead of post-1.0 nudge/drift work since they directly move retrieval-quality benchmarks.
codenamev
added a commit
that referenced
this pull request
Apr 30, 2026
`claude-memory show` prints what memory would inject at the next SessionStart. The trust answer to "is this still worth it?" requires inspectability — a user who can't see what memory will inject can't develop confidence in it. The CLAUDE.md alternative is `cat CLAUDE.md` (instant, plain English, no tooling); this is the same one-line inspect surface for the curated facts the injector picks each session. Runs the exact `Hook::ContextInjector` path real sessions use, so what shows here is what Claude actually receives — not a rebuilt approximation that could drift. Default suppresses the "Pending Knowledge Extraction" raw-transcript dump (intended for LLM distillation, not human reading); pass --pending to see the full fresh-session payload. --source SOURCE simulates startup/resume/clear so users can preview each entrypoint's payload. Footer reports fact count, ~token estimate (via Core::TokenEstimator), and char count so users instantly see the SessionStart cost — pairs with the #47 token-budget telemetry surfaces. Addresses: docs/1_0_punchlist.md (0.11.0 #5) / improvements.md #51
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Surfaces two ranking-signal gaps from Zaid's 2026-04-28 article "Why
Karpathy's Second Brain Breaks at Agent Scale":
factors into RRF ranking. Cheap win — data already exists.
and age decay don't influence retrieval order.
Both medium priority; sit behind the 1.0 punchlist but ahead of post-1.0
nudge/drift work since they directly move retrieval-quality benchmarks.