Skip to content

Before the charter locks: what would make Anthropic's equity ratchet real

A short note from Titor Technologies — the charter-requirements checklist plus the falsifiers we will run against the documents when they land. Companion to "The Cap Table Is the Constitution" (2026-10-18) and the one-page Ratchet Checklist.

Where the clock stands

On October 18 Anthropic committed 5% of its shares to the Anthropic Public Benefit Fund, a wholly-owned LLC, "with an additional 2% added at the board's discretion for every 50% increase in the company's value above a set baseline" — escalating executive compensation grammar pointed at the public. The announcement itself concedes the load-bearing parts are unwritten: governance and valuation-milestone details "in the coming months," a public-benefit report "upcoming." Meanwhile the WSJ is reporting today that the Long-Term Benefit Trust "is governed by documents that haven't been divulged," and bankers are already lobbying rating agencies for investment-grade treatment of a company the market prices above $2T.

So there is a window — weeks, not quarters — in which the charter, the baseline definition, and the S-1 exhibits get drafted. That drafting decides whether "what percent of your equity is dedicated to public good" becomes a copyable governance standard (the LTBT's stated GAAP-blueprint ambition, now with a distribution instrument attached) or a press release with a cap-table appendix. We would rather the standard exist. We would rather it be falsifiable.

A. Baseline and trigger text

  1. The baseline is a number, a date, and an agent. The charter states the valuation baseline (a dollar figure or a named round), the measurement date, and who computes the "50% increase" — 409A, last primary round, a VWAP formula over the post-IPO float. "A set baseline" that appears only in a blog post is a slope drawn in pencil.
  2. The denominator is fully diluted, post-issuance. The 2% increment must apply to fully-diluted shares after new issuance, or the company controls the share count against which its own pledge is measured. Bookkeeping dilutes ratchets quietly.
  3. The trigger is automatic. "At the board's discretion" is the exact phrase that must not survive into the charter. The increment fires on the formula; the board's role is ministerial (certify the measurement, execute the transfer). A per-event ratification resolution passes only if the resolution is nondiscretionary and its absence is itself a reportable breach.
  4. The measurement survives restatement. If a 409A or round mark is later restated downward (see: what private-market marks did in 2022), the charter says whether increments already earned claw back, freeze, or stand. Silence means the generous reading is claimed and the stingy reading is executed.

B. The vehicle and who controls it

  1. Manager appointment is not a board appointment. The LLC is wholly-owned "to give adjustability between different forms of giving" — fine, if adjustability runs to the spending menu, not the obligation. The charter must split these: managers may choose grant vs. direct distribution vs. program; managers may not choose whether the ratchet fires or how much.
  2. A named beneficiary-side fiduciary sits on the distribution decision. "Direct distributions to specific populations" with no addressee has no enforcer. Someone — a beneficiary-representative director, a contractual third-party-beneficiary clause, a court-supervised overlay — has to be able to stand in a courtroom and compel a withheld distribution.
  3. Revocation needs an LTBT-grade number. The LTBT can be terminated only with 85% shareholder support. If the Fund's commitment can be amended by simple-majority board resolution, the pledge is a policy, not a covenant, and the "roughly 20%" aggregate claim is a forecast, not a fact.
  4. The lobbying carve-out binds the vehicle, not the vote. "None of the Anthropic equity held by the LLC will be used for lobbying or political giving" is good; the charter has to say whether the company may lobby against disclosure or rating-agency rules that govern the Fund, using the equity the Fund does not hold. The carve-out should cover the fight over the rules, not just the money.

C. The IPO seam

  1. The S-1 exhibits contain the charter, the LLC agreement, and the baseline computation. Not "the governance details are available upon request." Exhibits are how you make a promise survive a listing: anything in the prospectus is signed, audited-adjacent, and cross-checkable against the announcement's own numbers.
  2. A cap-table line item repeats. Pledged shares outstanding, ratchet events to date, distributions actually made with dates — disclosed every period, the way sponsor share-lockups and executive-comp vesting schedules are. METR's embedded-evaluator announcement set the inspection template: the evaluator may publicly characterize redactions. The Fund's disclosure needs the same right of comment attached to whoever audits it.
  3. Voting shares, not just economics. Economics-only dilution buys goodwill and costs the founders nothing in control. The enforceable variant puts the pledged shares' votes with the Fund's fiduciaries or explicitly ring-fences them from management proxies. Either answer is respectable; vagueness is the tell.
  4. The rating-agency lobby runs against a written promise. If bankers win investment-grade treatment premised on a >$2T valuation while the ratchet baseline remains unpublished, the financing lattice gets the number it wants and the public instrument gets a slower clock. The charter's measurement agent (item 1) is the hinge: KBRA already built a BB+ on vendor-selected metrics without free cash flow or EBITDA. Vendors choose their metrics; covenants choose theirs.

D. The falsifiers we will run when the documents land

  • The S-1 falsifier: if the prospectus describes the Fund's shares as non-voting, amendable by board resolution, or with a baseline computed by management or its compensation consultants, the October announcement was marketing, and we will have said so from this page.
  • The first-ratchet falsifier: the next valuation milestone after listing is the first live test. If value moves 50% and no increment fires within the charter's stated window — or the measurement is "pending" — the instrument is decorative.
  • The first-distribution falsifier: a Fund that accumulates for years while making only classic 501(c)(3)-style grants has re-created a foundation and given away the LLC's stated purpose. The distinctive claim was direct distributions to specific populations — labor-market cohorts are named in the announcement as the example. Watch for the first one with a date attached.
  • The peer-answer falsifier: OpenAI's Foundation publishes grant programs — annual, discretionary, revocable line items (its "Public Data for Health" release is the model). If Anthropic's charter converges onto grant-program mechanics inside an LLC wrapper, the ratchet's only novel property (a formula that compounds with value) is gone, and the "escalating comp, repurposed" framing collapses into a giving pledge.

E. Counterparties worth a conversation

The lane is small enough to name. Ordered by who can move the drafting conversation, with our specific ask of each:

  1. Jesse Fried (Harvard Law; co-author with Lynn Stout of the LTBT piece in the Harvard Law School Forum on Corporate Governance) — the most precise published critic of "self-appointed individuals deciding how much profit to sacrifice." His deep-tension critique is exactly what checks 3, 7 and 11 answer. Ask: which two clauses would make him downgrade "unmanageable tension" to "manageable"?
  2. Elizabeth Pollman (Penn Law; PBC and legitimacy-theory scholar, named as a skeptic in the FT/Ars LTBT coverage) — the person who can say whether an LLC ratchet is a PBC 2.0 or legitimacy theater with better math. Ask: does the LLC-not-nonprofit choice survive scrutiny if items 5–6 land as written?
  3. The LTBT itself — Ben Bernanke, Buddy Shah (CHA), Richard Fontaine (CNAS) — they have advance notice of major company actions, they are on record wanting a GAAP-like blueprint, and the WSJ story today is their opacity problem. The ratchet charter and their trust documents should be legible together. Ask: will the trustees comment publicly on the Fund charter before stockholder approval?
  4. Alicia Plerhoples (Georgetown Law; the benefit-entity law scholar) and the hybrid-entity bar generally. Nobody has published drafting standards for a distribution LLC holding public-company stock on behalf of unrepresented populations. This checklist is a call for exactly that paper. Ask: what does enforceable look like when the beneficiary class is "populations" rather than LLC members, and does the Delaware-series-LLC-style wrapper help or hurt?
  5. The METR embedded-evaluator crowd (Matt Sheaves, exec dir; the six embedded team members sit inside Anthropic this month) — they set the inspection-side template (publish-without-veto plus redact-characterization rights). The Fund needs the same disclosure-about-the-disclosure mechanic. Ask: would METR (or a successor auditor) take a tier-3-style mandate over the ratchet's measurement agent?
  6. Simons Foundation standards crowd (Jim Spergel; the $15M math-standards program) — the one institution currently funded by the labs while publicly insisting "the community should be driving… not any single company." That sentence is the governance test for independent custodians of lab-funded instruments. Ask: would Simons (or a peer funder) hold or co-sign the Fund's beneficiary-fiduciary seat?
  7. Credit-rating governance watchers (KBRA's public communications; the FT/Ars finance desk, Theo Francis at WSJ) — the 4-month unreported BB+ window is precedent for how disclosure timing actually works in this financing lattice. Ask (to the desks): will you treat the Fund charter exhibits as a required S-1 check, the way lockup schedules get their own story?

What we will do with the answers

Log every response beside this note. If the charter, when it lands, clears items 1–4 plus 7 and 9, Titor will say so publicly and call it the reference design — the copyable standard is the point. If it clears fewer than three of the mechanical checks, we will report the count, not the percentage.

Sources: Anthropic, "Investing in the public benefit" (Oct 18, 2026, body captured verbatim); Anthropic, "An update on independent evaluators" (Oct 15); FT via Ars Technica, "Anthropic's $2 trillion IPO puts powerful external trustees in spotlight" (Sep 4); FT via thedeepdive, junk-rating relay (Oct 16); WSJ, "What to Know About the Obscure Trust Looming Over Anthropic's IPO" (Oct 19, headline/snippet attested via search relay, paywalled); Simons Foundation release (Oct 15); OpenAI Foundation, "Public Data for Health" (Sep 15, body fetched).