The working record behind the Foundry.
Why the Foundry is shaped this way, and how it is built — plus the prior-art surveys and verification reports that shape what gets built next.
Foundry design records
The core rationale: why the Foundry is shaped this way, and what each choice costs.
01 Positioning Statistical rails for LLM-driven statistical genomics — what the Foundry is, what it refuses to be, and why. READ THE RECORD 02 Guiding Principles The design pressure behind source authority, progressive disclosure, validation, and corpus grounding. READ THE RECORD 03 The Referee Loop How doing and refereeing compose — critique and calibration gating a result before it is trusted. READ THE RECORD 04 Architecture Map A short map of the Foundry's domain spine, implementation boundary, and focused architecture records. READ THE RECORD 05 Molds The axes Molds bucket on, the boundary against Pattern pages, and where the Mold set is still headed. READ THE RECORD 06 Mold Spec The Mold authoring contract — frontmatter, references, and the eval/scenarios companion shape. READ THE RECORD 07 Casting How typed Mold references become target-specific cast artifacts with provenance. READ THE RECORD 08 Corpus How external sources are ingested as own-words summaries without becoming a content mirror. READ THE RECORD
Project infrastructure
How it is built: the code, the content contract, what runs, and where files belong.
01 Code Architecture The current Astro and TypeScript implementation, dependency seams, entry points, and deliberate absences. READ THE RECORD 02 Content Model How statistical-genomics knowledge is represented as kinds, metadata, tags, references, and companions. READ THE RECORD 03 Build and Validation How the current corpus is validated, generated, rendered, and kept honest about deferred casting. READ THE RECORD 04 Repository Layout Where domain knowledge, implementation code, generated metadata, and source materials belong today. READ THE RECORD
Prior-art research
SOTA surveys and verification reports that map the landscape the Foundry is built against.
01 Double-Dipping Survey The review/survey layer above the primary-source notes — the double-dipping / selective-inference literature backing the double-dipping pattern. READ THE RESEARCH 02 Skill-Backing References For each cross-cutting bioSkill, the primary sources a traceable comparable skill recovers from — and how much is irreducibly convention. READ THE RESEARCH 03 Prior-Art Synthesis The cross-survey synthesis of the prior-art landscape for statistical rails over LLM-driven genomics. READ THE RESEARCH 04 Survey — Bioconductor / R State of the art through the Bioconductor / R lens. READ THE RESEARCH 05 Survey — Beyond R Statistical genomics prior art outside the R / Bioconductor world. READ THE RESEARCH 06 Survey — Genomics Broadly Reproducibility, QC, benchmarking, and truth sets across genomics broadly. READ THE RESEARCH 07 Survey — Statistics Broadly The referee at full generality — statistics broadly as prior art. READ THE RESEARCH
Related projects
Adjacent projects on their own terms — each one's framing and strengths, scored against the values the Foundry holds.
01 The AI Scientist Sakana's fully autonomous research loop whose reviewer self-certifies by reasoning — the referee anti-pattern; its framing, strengths, and where it sits on our values. READ THE PROJECT 02 Analyst-Inspector The closest architectural analog — an independent referee agent that gates reproducibility, not validity; its framing, strengths, and where it sits on our values. READ THE PROJECT 03 awesome-genomic-skills A discovery index above the genomics skill libraries (it catalogs them) — its framing, strengths, and where it sits on our values. READ THE PROJECT 04 BioinfoMCP Auto CLI→MCP wrapper for bioinformatics tools — the wiring layer beneath the tool call; its framing, strengths, and where it sits on our values. READ THE PROJECT 05 BioMANIA A natural-language → bioinformatics workflow agent that learns a Python tool's API and executes it — its framing, strengths, and where it sits on our values. READ THE PROJECT 06 Biomni A general-purpose biomedical AI agent — its framing, strengths, and where it sits on our values. READ THE PROJECT 07 bioSkills Our nearest neighbor on the repository-of-skills axis — its framing, strengths, and where it sits on our values. READ THE PROJECT 08 ClawBio A reproducibility- and benchmark-forward bioinformatics skill library on OpenClaw — its framing, strengths, and where it sits on our values. READ THE PROJECT 09 GenoTEX / GenoAgent An in-domain gene-expression benchmark + multi-agent system that self-certifies — its framing, strengths, and where it sits on our values. READ THE PROJECT 10 knowledgebase-mcp Agent-facing knowledge-retrieval MCP servers — their framing, strengths, and where they sit on our values. READ THE PROJECT 11 MCPmed A call to MCP-wrap bioinformatics web services for agent discovery/invocation — tool exposure, not a validity gate; its framing, strengths, and where it sits on our values. READ THE PROJECT 12 POPPER Our nearest neighbor on the referee axis — its framing, strengths, and where it sits on our values. READ THE PROJECT 13 StatReviewer + SciScore Deployed commercial automated manuscript referees — post-hoc, closed, reporting-level; their framing, strengths, and where they sit on our values. READ THE PROJECT 14 TusoAI An agent that designs and optimizes scientific methods against a metric — the upstream a validity gate sits beneath; its framing, strengths, and where it sits on our values. READ THE PROJECT