# E129+ Next-Evolution Research Roadmap

**Entity:** FIRME AI CORPORATION · **System:** CAIN-42 · **Baseline:** E128 (`d01f91a7e`)

1. **Authorized External Benchmark Integration (Priority 1):** Execute official ARC-AGI-2/3 public evaluation harnesses and METR/HCAST tasks under official rules with zero private-set contamination.
2. **Recruit Independent Human Baseliners & External Reviewers (Priority 2):** Convert `independent_external_reproductions` from 0 to >= 3 named external institutional/researcher reproductions.
3. **Scale Model-Driven Workers to Larger Local/Authorized Frontier Models (Priority 3):** Benchmark 7B–70B open-weights or explicitly budget-authorized frontier models across the full 2,000-trial campaign rather than 64 local 0.5B–1.5B trials.
4. **Non-Constructed Continual Learning & Real Software Repositories (Priority 4):** Move Worker 2 and Worker 4 from constructed discrete worlds into real multi-repo software engineering and natural-language continual adaptation.
5. **Production Gateway Reload, Live SSE & HF Space Sync (Priority 5):** With owner authorization, reload `cain-platform-gateway.service` to serve live `/api/v1/research/*` and SSE on all three sites, clean up legacy `Proto-VAGI` contradictions, install headless Chromium for automated visual mobile audits, and push updated HF Space static pages.
