AGENT_LEARNINGS.jsonl

{"_meta":"Episodic memory log. One JSON object per line. Append-only. No edits."}
{"ts":"2026-04-27T20:10:00Z","event":"memory_architecture_clarified","summary":"David clarified full memory taxonomy: working in WORKSPACE.md (volatile), episodic in AGENT_LEARNINGS.jsonl (append-only), semantic in LESSONS.md, personal TBC","pain_score":0,"tags":["memory","architecture","setup"]}
{"ts":"2026-04-27T20:13:00Z","event":"memory_taxonomy_fully_defined","summary":"David provided complete memory taxonomy: WORKSPACE.md for volatile working memory, AGENT_LEARNINGS.jsonl for append-only episodic, LESSONS.md for semantic, personal/ directory for personalprefs. Clear rules given for each type. Personal memory location still TBC.","pain_score":0,"tags":["memory","architecture","taxonomy"]}
{"ts":"2026-04-27T20:15:00Z","event":"semantic_memory_write_rule_clarified","summary":"LESSONS.md is read-only. May propose via citing episodic entries + durable rationale + draft. Requires explicit approval before any write. No direct writes.","pain_score":0,"tags":["memory","semantic","LESSONS.md","protocol"]}
{"ts":"2026-04-27T20:22:00Z","event":"semantic_lesson_proposal_rules_established","summary":"David defined explicit triggers for proposing semantic lessons: recurring pattern (2+ entries, 30 days), high-pain/high-importance (>=8/>=7), architecture/permissions/control, or explicit request. Also defined explicit exclusions.","pain_score":0,"tags":["memory","semantic","LESSONS.md","protocol","proposal_rules"]}
{"ts":"2026-04-27T20:25:00Z","event":"semantic_memory_limits_established","summary":"LESSONS.md hard limits: max 30 lessons, max 200 lines, single sentence each, max 1 per 10-15 episodic entries, max 1 per week. Overlap/contradiction triggers promotion pause until consolidation. Log and explain if limit hit.","pain_score":0,"tags":["memory","semantic","LESSONS.md","limits","protocol"]}