Early-stage funds holding AI stock: when a markup round arrives with no public revenue or fee data since the last pricing, the second price is first a window to sell part of the position inside that round, because the next pricing after an operating cycle may not hold it. Instinct was valued at $2.5B on Aug 26 and $10B on Sep 28, quadrupling in 33 days (author's calculation from announcement dates); the funding announcement disclosed no revenue or user numbers, the founder self-reports annualized transaction volume of nearly $1B (via TechCrunch, Sep 29), and a16z (Oct 5, the first day of SF Tech Week) says no transaction fees are charged yet. The counter-case: Benchmark invested in both the B and C rounds.
Investors: in such a round within 90 days of the last pricing, GPs should ask for room to sell part of their stake as secondary inside the round; LPs should ask GPs how much they sold in such rounds over the past 12 months, require a written reason where they did not, and weigh re-ups by the answer.
Founders: add a secondary tranche alongside new shares so the new lead fills its ownership target from existing holders; don't trade liquidation multiples or ratchets for a headline valuation. Sell or hold turns on whether an operating cycle sits between the two prices, not on the paper multiple. Falsifier (counted from Oct 11, 2026): a single-case test on Instinct against the $10B headline valuation of Sep 28, 2026 (pre- or post-money undisclosed); the test uses 180 days after the Series C as a proxy for an operating cycle, whether or not revenue or fee data is published in between. If the first pricing announced or reported between Mar 27 and Dec 31, 2027 (a primary round, or a company-run secondary or tender; a pricing reported by only one outlet does not count as the first), or any M&A, IPO or license-and-hire at any time from Oct 11, 2026, comes at or above $10B with no reported over-1x or participating preference, valuation or IPO ratchet, or guaranteed return, the call is wrong; a lower price, any such term, or a lower deal value verifies it; ⏳ until then, final check Jan 14, 2028; with no event by Dec 31, 2027 the window extends to Jun 30, 2028, and an event in the extension is judged the same way with a final check on Jul 14, 2028; if there is still none, it is recorded as 'source missing: no pricing event in the period', neither wrong nor verified. Pricings before Mar 27, 2027 are recorded, not judged. Whichever qualifying event comes first decides; later events are recorded only.
Disclosure: written with Claude; Instinct has not named its model providers, and its consumer agent competes with Claude's consumer products (an inference); Series C investors Sequoia and Coatue, and seed investor Greenoaks (per secondhand reports), are Anthropic investors, so discount both ways; FutureX holds no direct or indirect stake in Instinct or Anthropic, including through SPVs, funds of funds or LP interests in other funds, among them funds of Instinct's named investors Sequoia, Benchmark, Coatue and Index and of Greenoaks, Kleiner and Conviction, reported secondhand as early-round investors (confirmed by the author from FutureX's records); FutureX holds Genspark, an Anthropic customer in the same agent category, so the author has an interest in the operating-cycle definition, and this entry does not say which side of the rule Genspark falls on; Corgi and Mistral, named in the footnote, are also FutureX holdings, left out of the sample without being classified, a choice in FutureX's interest; FutureX holdings ByteDance (CapCut, Doubao and others), Hugging Face and Pixocial (BeautyPlus) appear in the a16z Top 100 cited here, and Hugging Face has a pending acquisition by Nvidia announced on Sep 3, 2026, which aligns with this entry's M&A counter-argument, so discount that too; this entry does not pre-announce any FutureX sale or valuation.
This diary does not constitute promotion of, or an offer to subscribe for, any fund product.