method

the lanes

the corpus was produced by two parallel agent research lanes — one covering company history and the people, one covering coverage, sentiment, and the automated-lab idea — each writing a lane report against one shared scratch source catalog before synthesis. the pattern is the company-research formula proven on earlier dossiers: lane-prefixed source ids frozen upfront, per-claim confidence, and a declared deduplication owner.

the evidence rules

1. every event, quote, and claim carries at least one source id.
2. dates keep their precision — a year, a month, or a day — and approximate dates are labeled rather than sharpened.
3. confidence stays honest: confirmed, reported, and inferred are distinct claims, not stages of the same claim.
4. sources are typed by role — primary, interview, reporting, archive, community, reference, repo — so a claim's foundation is visible before it is read.
5. sentiment labels describe the coverage, not the truth of it.

known gaps

1. no publicly documented commercial product, api, or pricing exists as of september 2026 — the near-term "product" is the lab's own research loop.
2. the funding numbers never reconciled in public: ~$100m reported on safes at ~$1b; a reported $300m–$500m seek at ~$4b; the filing shows $635.2m authorized, $432.1m sold, and $135.3m of safe conversion that doesn't match the reported seed size. the discrepancies are preserved, not averaged.
3. the form d's 51 investors are unnamed — sequoia and acrew positions are inferred from a hosted podcast and a board seat, not disclosed.
4. there is no official team page — the roster is assembled from self-applied linkedin titles, launch posts, and the filing; swapnil patil's identification rests on a name match, and one aggregator's "adept cofounder" claim for anmol gulati is likely a conflation.
5. the reported "ceres" model — single-phase, ~100× less data, continuous deployment learning — is second-hand through the information's relays; the company has never confirmed it.
6. the two anchor information pieces are paywalled; the record runs through techmeme, dealroom, and the decoder's summaries.
7. the x corpus is partial — posts came through archives and mirrors, not native search; the company's own account has posted little since launch.
8. headcounts disagree — 14, ~21, and "20-something" all appear in sources; each keeps its provenance.

corrections

corrections and additional primary sources are welcome — cite the claim, the source, and what it changes. the catalog is designed to be audited, not just read.