CognitiveHire -- Divergent Rail Execution Plan

> Generated: 2026-03-27 > Protocol: Divergent Rail (EW-governed parallel execution) > Project: `[home]/Desktop/cognitive-hire/` > North Star: Mohamed Diomande -- 112K+ AI turns, KARL adapter, RAG++ 332K rows | Source | Location | Size | Format | |--------|----------|------|--------| | Prompt logs | `[home-path]` | 3.4GB, 6099 lines | JSONL | | Supabase prompts | `claude_prompts` table | 112K+ rows | SQL (prompt_text, tags, complexity_score, intent_classification, tokens, timing, git context) | | Supabase turns | `claude_assistant_turns` table | N x turns per prompt | SQL (content_text, thinking_text, tool_calls, model_id) | | Supabase tools | `claude_tool_calls` table | tool_name, tool_input, duration_ms, success | SQL | | Supabase diffs | `claude_file_diffs` table | file_path, action, lines_added/removed, unified_diff | SQL | | Supabase daily | `claude_daily_summaries` table | Materialized daily stats | SQL | | RAG++ embeddings | `memory_turns` table | 332K rows, 768-dim (text-embedding-3-small) | pgvector | | KARL trajectories | `[home-path]` | 121+ trajectories, 72 skill-labeled, 11 domains | JSON/PKL | | KARL config | `[home-path]` | Embedding: gemini-embedding-001 3072-dim via RAG++ :8000 | JSON | | Prompt logger | `[home-path]` | 91 files, analytics.py, mcp_server.py (36 tools) | Python | > **Gate**: Raw cognitive metrics computed for Mohamed. At least 3 metric types producing non-trivial output from real data. Dashboard skeleton renders locally. > **EW Check**: Invariant 1 (Minimum Entropy). Risk: metrics produce flat/meaningless output from real data. Corrective: swap metric algorithm, not data source. | Track | Task | Machine | Blocks | EW Role | |-------|------|---------|--------|---------| | Critical | Cognitive metric extraction scripts | Mac1 | Phase 2 (dashboard needs data) | Rail spine | | A | Dashboard scaffold (Next.js + D3/Recharts) | Mac1 | Phase 2 Track Critical (renders metrics) | Divergent organism | | B | ChatGPT/Gemini export parsers | Mac1 | Phase 3 Track A (multi-source ingestion) | Divergent organism | | C | Content: "Today vs Tomorrow" script draft | Mac1 | Phase 4 Track B (video production) | Divergent organism | Build Python scripts that query Supabase directly and compute 6 cognitive metrics from Mohamed's real data.