Bactaegion is a demo of what AI well-used can produce in open science.

This platform was prototyped in a few weeks with Claude (Anthropic) as code partner, on a real case: the bacterial defense atlas the Institut Pasteur published in May 2026. The result is not a finished product, it's a prototype — see the public roadmap for the exact state of each piece.

Can we, in 2026, build a credible open-science platform in a few weeks rather than several years?

With a quality LLM as co-pilot (spec reading, code generation, critical audit, propositional design, refactor, sub-agent tests), the answer is yes — provided a rigorous human stays in product steering to: validate scientific choices, course-correct when the machine produces wishful thinking, demand honesty on undelivered promises.

• Reading and synthesis of the two preliminary Gemini reports (TAZ architecture + translational audit).
• Astro 5 + React 19 + strict TypeScript implementation (all v4 components).
• Iterative redesign of screens in response to product critiques (cinematic, triptych, gestures, dossiers).
• Generation of calibrated scientific content (12 piano-roll scenarios, 6 annotation proteins, peer-review hypotheses) from named public sources (Bernheim 2026, Doron 2018, etc.).
• Continuous critical audit — sub-agents that test the paths and flag regressions.

• Product vision: what Bactaegion is for, who for, against what (see /en/mission/).
• Scientific framing: choice of priority therapeutic targets, distinction between realistic pipeline and fantasy (therapeutic CRISPR-Cas = excluded, Viperins + CBASS + Pycsar + Schlafen + RADAR + Thoeris = retained).
• Game design: refusal of dark patterns (XP, leaderboards, loot boxes, push), choice of permanent epic sense instead.
• Honesty watch: requiring that everything said on the site exactly match what is shipped. No Open Badges / DID / ORCID / Yjs announced as imminent if they are not in code (see /en/roadmap/).
• Translational strategy: identifying the relays needed outside Bactaegion (DNDi, GARDP, M4K Pharma) to bring a hit to clinical.

We could silence the massive use of AI in building Bactaegion and pass for a larger human team than we are. We choose the opposite: openly declare the production method. If the platform stands scientifically and works as an open-science prototype, then it proves something useful about what can be done in 2026 with AI well-used. That's its meta raison d'être.