Skip to content

Roadmap

The factory and this site grow together, in public. The Capability Maturity Matrix is the live scorecard; this page is the honest direction.

The site goes green as the factory matures — roadmap to planned to live to verified

Every capability advertises its real status, generated from the registry — the site cannot claim ahead of the code.

Platform

  • Live today — the platform library (capability registry, MCP tool-surface, workflow composer, governance gates) and the engines/agents the registry declares live.
  • Planned — the frontier co-folding and generation models registered planned (e.g. Chai-1, GenMol) flip to live only when they are stood up and served for real. (Chai-2 de-novo binder design is gated — partnership access — not planned.)
  • Preclinical — gene-therapy directions (including TSC1/TSC2) are an open design/analysis bench, not a treatment available today.

This site

Built in four phases, each flipping matrix badges green as real capabilities come online:

Phase Focus
0 — Foundation (shipped) design-system + honest status badges, OG/SEO + llms.txt, the registry-bound maturity matrix, the section skeleton, a strict docs build in CI
1 — Depth (shipped) per-engine / per-agent / TSC pages, published diagrams + narrated videos, badges flipping as capabilities verify
2 — Proof (shipped) "Run It Yourself" quickstart, verified-on-real-data proofs, citations, governance
3 — Launch (in progress) performance + accessibility pass, analytics for the right metrics, the launch moment

Changelog

The site build, in public — most-recent first.

  • 2026-07 — Verification campaign: ran capabilities on real public data and earned verified where they passed — nine in total: the variant store (HG002), single-cell compute (pbmc3k), ACMG secondary-findings (real ClinVar), the molecule generator + ADMET (real drugs), ESM-2 search + ProteinMPNN + ESMFold (real proteins), and the reproducibility capture.
  • 2026-07 — Fuller IA: two-tier tab navigation, a page for every engine and agent, About / Mission, Honesty Ledger + Decision-Support Posture pages, Run It Yourself split (Quickstart / Hardware / Reproducibility), per-page social OG cards, CITATION.cff.
  • 2026-07 — Honesty audit: code-level review of every live capability; corrected the real overclaims (live → planned where a dependency can't load); the maturity vocabulary shipped.
  • 2026-07 — Foundation (Phases 0–3): the registry-bound maturity matrix, the honesty spine, citations & governance pages, llms.txt + SEO, the design-system status badges, a strict docs build gating every PR.

A note on the status vocabulary

Beyond the serving states live and planned, the registry now carries an optional, orthogonal maturity tier — verified · preclinical · research-use · roadmap · gated — machine-checked against the code (e.g. verified is rejected by the registry on anything not actually running). It is populated conservatively, only where the registry documents a basis: today nine capabilities are verified (proven on real public data — the variant store on HG002, single-cell compute on pbmc3k, ESMFold, ESM-2 search, ProteinMPNN, ACMG secondary findings, the molecule generator, ADMET, and reproducibility), and Chai-2 is gated (partnership access). Broadening coverage — and adding finer-grained capability entries for the feature-level caveats now carried on the Honesty & Governance ledger (e.g. TSC gene therapy is preclinical) — continues as capabilities are audited.