Skip to main content
Bactaegion
EN · FR
✦ Public roadmap

What ships, what doesn't, what depends on partners.

Bactaegion is today a consultable prototype end-to-end. Not a finished product. This page honestly lists what is shipped, what is being shipped, and what has not yet started. No feature is announced as imminent if it hasn't started.

Raw format: one deliverable = one line. Status, commit hash, done criterion. Canonical source: ROADMAP.md in the repo.

shipped and working 🟡 shipped but incomplete/mock 🚧 in progress not started out of B0 (conditional)
J0

Consultable prototype

✅ shipped

Current state on bactaegion.pages.dev. Cinematic, triptych, 3 orientation paths, 3 gestures (piano-roll, annotation, peer-review), cabinet with 14 dossiers, accessible BYOK LLM, refactored canonical sources, language picker forced for new visitor.

Honest limits: the mission counter («0 / 478,206 families touched») reads your localStorage — that's what YOU did, not the community. The «community consensus» mock detects only the other tabs of your browser. No real collective aggregate exists.

J1

Honesty + B0 polish

🚧 in progress

Goal: everything said must exactly match what's done. Creation of this roadmap page + mission page, purge of technical mentions («Open Badges 3.0 / DID / ORCID / Yjs / W3C») in user surfaces as long as nothing is implemented, displacement of BYOK LLM outside the entry paths, contextual banner on each action page recalling which objective is served, refactor of the cinematic's dramatic pacing, navigation tests of the paths.

J2

B0 corpus densification

⬜ not started

Enrich the 11 cabinet dossiers currently in «research tier» to move them to «curated tier» (add protein_ids, Wikidata qid, Wikipedia URL). Complete the remaining 14 dossiers (1-15 to 1-28). Add +3 piano-roll scenarios (reach 15), +6 annotation proteins (12), +5 peer-review hypotheses (10) on the priority NS-2 targets. Mention «Claude Anthropic AI demo» in the footer + page /comment-on-a-fait/.

J3

First real scientific cycle (B0 → B1)

⬜ not started

Transition from tutorial to real contribution. Ingestion of the GeneCLRDF Bernheim 2026 dataset (~10,000 orphan families). «Real contribution» mode on piano-roll and annotation (choice between calibrated cases and unvalidated families). Minimal aggregation backend (Cloudflare Worker + Turso SQLite, zero-cost TAZ doctrine). EpicCounter switches to global mode. Automated GitHub PR export from BYOK LLM. Arweave timestamping of CC0 hypotheses to create a prior-art shield against patent capture.

J4

Decentralized identity layer (V4.1)

⬜ not started

Browser-side did:key generation via WebCrypto. Local signing of contributions (annotations, hypotheses). Server verification before inclusion in the aggregate. Open Badges 3.0 (W3C Verifiable Credentials) emission tied to the DID, persistence via the DID. Optional OAuth ORCID via Cloudflare Worker to anchor the DID to a verifiable academic profile.

J5

Yjs CRDT real-time collaboration (V4.1.5)

⬜ not started

Collaborative editing of annotations via Yjs CRDT. Real-time aggregate of peer-review verdicts. WebRTC transport via self-hosted Hocuspocus or browser libp2p, depending on stability observed at Cloudflare.

J6

Federated Learning (V4.2+)

❌ out of B0 — depends on partners

DAP/Daphne client in the browser. Cloudflare Worker as Leader + Helper. Depends on recruiting at least one laboratory partner ready to train locally on private data. As much strategic as technical work. First federated model trained on ≥2 nodes.

Off-roadmap — translational

Clinical pickup of hits produced by Bactaegion depends on non-commercial sponsors (DNDi, GARDP, Wellcome Trust, M4K Pharma). This is strategic work, not code work. It will be announced separately as agreements are put in place.

↗ see the full mission ↗ back home