Open
Conversation
… homeostasis verification
Three council findings addressed with architectural changes:
1. POPPER (falsification): Verified homeostasis causal chain honestly.
Quality gate DID use compass truthfulness to tighten thresholds,
but compass data came from self-reports + auto correction_rate,
not from pull detection auto-feed (which didn't exist yet).
Homeostasis real but bootstrapped by honest self-reporting.
2. SCHNEIER (attack tree): Added 4 subtle epistemic markers that
close the barn-sized blind spot in pull detection:
- unwarranted_certainty (3+ confidence words without evidence)
- unmeasured_claim (comparisons without numbers)
- evidence_gap (self-check marker for assertions without sourcing)
- false_consensus ("research shows" without citing research)
Soft markers warn + feed compass at -0.15 (vs -0.40 for loud).
They accumulate gradually — death by a thousand cuts detection.
3. GODEL (incompleteness): HUD self-awareness slot now shows
verification gap when all compass data is self-assessed with
zero user ratings. Makes structural self-verification limits
visible rather than hiding them.
Also includes from previous session:
- Pull detection system (6 loud fabrication markers)
- RT protocol with pull gate enforcement
- Engagement gate decay (light tools halve, deep tools reset)
- Council expert templates (28) brought into worktree
- 30 pull detection tests (all passing)
- Doc counts updated (151 commands, 3,699+ tests)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…core memory bugs Three root causes fixed: 1. Quality gate false blocking: check_correctness scored 0.0 when test results existed but were filtered out by since_timestamp (context compaction hides earlier context windows). Fixed with: score changed from 0.0 to 0.3 for missing-tests-in-coding-session, plus expanded window fallback that retries without timestamp filter. 2. Compass blind spots: 6 of 10 spectrums were starved because their auto-feed thresholds only triggered on failure signals. Normal productive sessions produced zero observations. Added baseline observations for thoroughness, confidence, compliance, precision, and humility that fire during healthy sessions. 3. Core memory refresh bugs: _refresh_strengths() used a closed DB connection (conn.close() on line 89, reused on line 108). Fixed with separate connection. Also removed unused get_knowledge import. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
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
1 participant
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.
No description provided.