docs(specs): high-value implementation plan proposal from .specs/ review

[glassBead-tc]

Summary

Adds .specs/PROPOSAL-HIGH-VALUE-IMPLEMENTATION-PLAN.md: a prioritization-and-sequencing proposal produced by reviewing the entire .specs/ corpus (171 files across ~23 clusters) and grounding every idea against two reality anchors — PRODUCT-INTENT-AND-DIVERGENCE.md (settled intent) and PRODUCTION-SYSTEM-MAP.md (what is actually live).

The document identifies the highest-value designs and proposes how/when to build them, with explicit reasoning for every choice (as requested).

Method

Ranking axes are stated up front so the prioritization is auditable: intent alignment, leverage (value × proximity to done), unblocking power, risk/blast-radius, and falsifiability. Load-bearing status claims were verified against current source rather than trusting spec metadata (e.g. confirmed the tb SDK has no tb.hub, the public surface is 3 tools, Hub storage is FS-only).

The eight ranked plans

1. Make the Hub real on the public surface — durable SupabaseHubStorage → tb.hub in Code Mode → Hub HTTP auth. The flagship is ~80% built but invisible; highest leverage.
2. Consolidate the Code Mode public surface — kill documented drift (tb.hub/thoughtbox_operations/init/cipher), fold guidance into search.
3. Core thought-tool ergonomics + mid-session recall — surface auto-numbering, tb.t(), recall primitives, named checkpoints.
4. Causal evaluation loop — four-condition ablation + diagnostic metrics on the existing src/evaluation/ code; proves the product claim.
5. Make the reasoning graph visible — Knowledge Graph UI + auditability rendering in the web app.
6. Notebooks as executable evidence — finish evidence engine + peer-notebook manifest lifecycle; merges that carry proof.
7. Ship-hardening — runs correlation, frictionless init, explicit failure states (per USABILITY-CUTLINE).
8. Control-plane learning (frontier, thin slices only) — sleep-time evolution-check, Tier A gates, tool pedagogy; DGM/MAP-Elites kept as research.

Also includes a proposed sequencing, a "what I deprioritized and why" section, and open intent questions that gate the flagship plans.

Notes

• This is a proposal document only — no code changes. It points back to the authoritative specs that should each go through the HDD lifecycle (.adr/staging/ → .adr/accepted/) before implementation.
• All cited src/ and supabase/ paths were verified to exist.

Testing

• ⚠️ No automated tests apply (documentation-only change). Pre-commit hooks ran: ✅ cyclic-dependency check, ✅ oxlint (0 warnings/0 errors).
• ✅ Verified every cited file path exists (src/code-mode/execute-tool.ts, src/hub/*, src/branch/*, supabase/functions/tb-branch/, supabase/functions/process-thought-queue/, src/audit/manifest-generator.ts, etc.).

  

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
thoughtbox-web-two	Ready	Preview, Comment	Jun 3, 2026 5:22am

[supabase]

This pull request has been ignored for the connected project akjccuoncxlvrrtkvtno because there are no changes detected in supabase directory. You can change this behaviour in Project Integrations Settings ↗︎.

---

Preview Branches by Supabase.
Learn more about Supabase Branching ↗︎.