Pramana Labs

What we have established three decisions, kept apart

1. 1 Identity verified Yes

   CLAIM BY CLAIM, because these claims are not equally supported (J, 2026-10-11). SUPPORTED: the operating business at pramaanalabs.ai is the company the transcript described — its own page states the $27M raise that coverage attributes to Pramaana Labs, and its subject matter (formal verification, Lean, tax) matches both the transcript and the coverage. ALSO SUPPORTED: the GitHub organisation pramaana-labs is ASSOCIATED with that site, because it declares the site as its blog and its repositories are on the same subject. NOT ESTABLISHED: that the organisation is owned or controlled by the legal entity — a GitHub account can name any website. NOT ESTABLISHED EITHER: the legal entity itself. We hold no filing, no registration and no incorporation record; "Pramaana Labs, Palo Alto, founded 2025" is press description, not a primary identity document — which is a weaker identity position than any of the six Form D companies, every one of which was matched to a filing by person-level identifiers.

   • Transcript (2026-10-05 and 2026-10-09, two different shows): "Pramana Labs would be one of them that is going after taxes" / "PMAN labs… going after taxes"
   • Public coverage spells it PRAMAANA Labs: $27M seed, Khosla Ventures, Palo Alto, founded 2025
   • pramaana.ai is a parked for-sale page and pramaanalabs.com is a 114-byte shell; the GitHub organisation pramaana-labs declares blog = pramaanalabs.ai
   • pramaanalabs.ai: "Pramaana Labs - AI That Proves Its Work… We Raised $27M To Build A Compiler For Mission-Critical AI"
   • Its repos are Lean4 (catala-lean: "literate programming law specification"), matching the coverage’s description of formalising the tax code
2. 2 Evidence sufficient to publish Not yet

   The inclusion bar: two or more independent sourced evidence families. This profile is reachable but not in the ranked index. It is a statement about how much evidence we hold, and about nothing else.

3. 3 Fit for a seed/Series-A sourcing buyer Qualified

   Institutional venture backing and a stage band inside seed..Series A are both evidenced below.

   Both legs hold. (a) Khosla Ventures LED a $27M round; early backers include Pushmeet Kohli (Google DeepMind) and Sriram Rajamani (Microsoft Research). (b) Every source calls it a SEED, dated June 2026, and it is the only financing event on record. Formal verification over regulated domains, starting with the US tax code, with a former IRS Commissioner advising. Palo Alto, founded 2025.

   Venture backing: evidenced · Stage band: seed — stated by coverage of the round, June 2026

   • dealstreetasia.com ↗ — $27M seed led by Khosla Ventures.
   • pramaanalabs.ai ↗ — The company’s own site, confirming the $27M and the business.
   • github.com ↗ — The GitHub organisation whose blog field pointed at the real domain. Lean4 repos (catala-lean, imo2026-lean). NOT attached as developer evidence — see the unresolved question.

   What the filing alone says

   NOTHING — no SEC filing is in evidence. This company came from two sentences in two different podcasts, five days apart, both mis-spelling its name.

   What we add

   The hardest identity resolution in either pass. Neither transcript spelling ("Pramana Labs", "PMAN labs") finds the company; the obvious domain is parked for sale; and the working domain was reached only through its GitHub organisation’s own blog field. On top of that: the $27M Khosla seed, and the Lean4 repos that corroborate the business the transcript described.

   Still unknown

   • Its $27M seed gives it ONE sourced evidence family, so it is not publishable. The GitHub organisation would be the second — see the blocker below.
   • Whether the performance claim on its own site ("never produced a confidently wrong verified answer") has been independently tested. We found no such test.
   • Which other investors joined; the company says several firms did, unnamed.

   Action it supports

   A qualified name a buyer did not have, with the founders, the lead, the stage and the technical substance attached. Not yet a profile page, which is a limitation of our evidence rule and not of the company.

   Our arrival vs public

   ~4 months LATE on the round (June 2026; entity created 2026-10-11T03:43:04Z).

   Research time

   see the Aristotle entry: ~13 minutes of machine time for all three.

   Unresolved

   • THE PUBLICATION BLOCKER, stated exactly: github.com/pramaana-labs is named "Pramaana Labs", declares blog = pramaanalabs.ai (the verified site) and hosts Lean4 tax-formalisation repos. The existing rule in resolve_github_repos.ts asks for OWNERSHIP EVIDENCED FROM THE COMPANY’S OWN PAGE — the account named twice, or named once with a backlink. pramaanalabs.ai names no GitHub account at all, so the rule is not satisfied. (It is not a "bidirectional link requirement", which is how I described it on 2026-10-11; a single mention plus a backlink passes. The requirement is that the company’s own page mention the account once.) Association is supported; ownership is not; the rule was neither relaxed nor waived.
   • Whether the "never confidently wrong" claim survives independent testing.

   What would overturn this: Evidence the Khosla round was not a seed, or that the company is outside the buyer’s mandate.

   Assessed 2026-10-11 for seed/Series-A sourcing.