90th Percentile Hits $99.8M: The 90th percentile of seed post-money valuations for US companies on Carta reached $99.8M in Q2 2026, up from $20.0M in Q1 2018, a roughly 5x increase over the period.

Median Rises to $27.0M: Median seed post-money valuation climbed to $27.0M in Q2 2026 from $9.2M in Q1 2018, a rise of nearly 3x over the same stretch, a far slower climb than the 90th percentile.

90th-to-Median Gap Widens to 3.7x: The multiple between the 90th percentile and the median grew from 2.2x in Q1 2018 to 3.7x in Q2 2026, the widest spread in the dataset.

AGI's Two-Question Framework: Acharya splits AI progress into capability, whether the model can do the task, and economic impact, whether organizations can absorb the resulting workflow change.

Closed Models Price 4.6x Above Open Models: Silicon Data's LLM token price indices put the closed-model index at $3.07 versus $0.66 for open models; Acharya frames this gap as a routing signal, with real differentiation still sitting at the application layer.

Moats Persist Selectively as Agents Spread: Network effects, distribution, brand, and scale still favor incumbents, Acharya argues, but the edge from linking software together is fading as agents learn to do it themselves.

Two in Three Checks Lose Money: A Correlation Ventures census of about 21,000 US financings (2004-2014) puts the loss-or-flat rate at 65.3%, while just 0.53% of deals land at 50x or higher.

Harry Stebbings Couldn't Predict His Own Fund's Winners: After the deployment period for 20VC Fund I, Stebbings ranked his predicted top 5 portfolio companies; three years later, not one of them matched the fund's actual top performers.

Overconfidence Fuels Overconcentration: Fund managers who believe they can consistently pick winners tend to overconcentrate on favorites; a 2022 paper found this backfires more than it pays off, hurting weak funds more than it helps strong ones.

Sell the Hammered Nail, Not the Hammer: The idea behind FDE (forward-deployed engineering) is selling the hammered nail: a company guarantees the finished business outcome itself, backed by hands-on service, and prices that guarantee directly.

AI Rollups Turn Service Firms Into the Product: In this rollup model, a firm buys existing service businesses and uses AI to compress their costs; Slow's Metropolis used this approach to roll up parking and is now valued at $5B.

Agent Networks Flip From Labor Into a Network: Rechtman's agent-network model starts as paid labor, like Recurrence's Phoebe filling caregiver shifts, then resells the worker network it built to new demand, an end state he calls a "for-profit union."

Great Hires Are an Underused Signal: Few founders showcase recent hires in their pitch, per Kothari, even though strong hires help VCs underwrite the downside, since a struggling company can still be acquired for its talent.

Obviousness Is the New Series A Bar: No revenue figure guarantees a Series A anymore, a point Kothari draws from Yoni Rechtman; the real test is whether the opportunity looks obvious, and justifying its size to investors is already a bad sign.

Word Travels Fast in a Small Market: Claiming a term sheet from one fund often reaches competitors within a day in a small, connected market, Kothari notes, so founders should assume anything they say will circulate.

Feel Like an Owner, Don't Burn the Ships: Drory argues founders should protect the sense that their success and the company's are the same, while paying themselves enough to avoid debt or early burnout, since underpaid founders tend to quit before the real outcome arrives.

Don't Overdo It, Either: Drory warns against living like a "wantrepreneur" chasing status over risk; what a founder pays themselves also sets a soft ceiling on what they can justify paying senior hires.

Four Triggers for a Comp Change: A material life change (kids, an emergency, or real risk of personal debt), dilution down to employee-level equity, the company reaching escape velocity, or a well-timed secondary at a high, justified valuation.