Key Takeaways

  • Benchmark backed 23-year-old Noah Shin's startup Instinct at a $2.5 billion valuation and followed on at $10 billion, treating both checks as early-stage risk.
  • Instinct raised a $1 billion Series C to build consumer agents that act as an aggregator of aggregators across the third-party web.
  • Rory O'Driscoll argues the current AI market forces funds to deploy late-stage capital into bets that still carry raw product-market fit risk.
  • Jack Altman admits venture pricing models are broken right now: firms are either moving far too quickly or missing the market entirely.

The Disagreement

Traditional venture math says a $10 billion valuation belongs to companies with hundreds of millions in predictable ARR and proven retention curves. Instinct blew past that standard. Founder Noah Shin raised a $1 billion Series C at a $10 billion valuation to build autonomous consumer software.

Benchmark leaned in heavily. Altman explained the firm's logic: “And we invested first at 2 and a half and 10. And what's funny is in our minds it was actually kind of an early stage investment.” To Altman, Instinct represents consumer software agents that interact directly across the entire third-party internet. If Instinct succeeds in becoming the layer between consumers and every other web service, conventional revenue multiples do not matter.

O'Driscoll sees a structural hazard in this logic. He points out that venture funds are now underwriting seed-level uncertainty with late-stage balance sheets. “In this cycle we've had investments with early stage risk requires super late stage capital which is just definitionally a strange time to be playing,” O'Driscoll said. When a company with unproven long-term usage raises billions, any slip in product adoption wipes out investor returns because there is no safety margin left in the price.

Lemkin frames the outcome on pure daily screen time: “Will we run Instinct Muse 8 hours a day, right? If we do, I guarantee it wins, right?”

Altman does not pretend the math makes neat spreadsheet sense. “We are either investing way too fast or way too slow,” he said. “But when both supply, you know, when both sides of the equation are this out of whack, the odds of having it right are zero.”

Who's Right (and When They're Wrong)

Benchmark is right if consumer agent behavior creates a winner-take-all monopoly on web traffic. In that scenario, paying a $10 billion entry price is cheap because the winning agent captures trillions in downstream commerce. The classic rules of software multiples fail when an interface captures aggregate consumer demand.

O'Driscoll is right for almost every other firm in tech. When you pay a $10 billion valuation on an early product, you need a $100 billion outcome just to return a modest multiple on that specific check. Unless your fund has the exact size, reserve pool, and tolerance for binary wipeouts that Benchmark commands, pricing seed-stage execution risk at sovereign-fund scale is a direct path to destroying your fund metrics. If users do not keep Instinct active all day, the capital disappears.

What to Do With This

Audit your company's product defensibility against third-party agent layers. Map every customer workflow in your product and ask: if an autonomous browser agent can complete this entire task via external APIs in 30 seconds, why would a human ever log into our web app again? If your moat depends on UI friction, start rebuilding around proprietary data assets that agents must pay to query.