Sequoia's David Cahn is backing Flapping Airplanes, a research lab founded by Ben and Asher Spector, ages 25 and 26, alongside Thiel Fellow and ex-Neuralink engineer Aidan Smith. The thesis is direct: data, not compute, is the bottleneck to the next phase of AI scaling. The internet has been exhausted as a training source, and today's models are data-inefficient by any biological standard. Cahn cites Dwarkesh Patel's recent interviews with Andrej Karpathy and Richard Sutton as the clearest public articulation of this problem.
The lab is structured as a deliberate counter to the current industry default of pouring capital into cluster scale and hiring brand-name researchers at any price. Flapping Airplanes offers research independence and long time horizons, modeled on a Ph.D. experience but with competitive compensation. Cahn argues that funneling raw talent into scale-oriented work, rather than fundamental research, could actually extend the AGI timeline, not shorten it. That argument, and the tradeoffs between compute-first and research-first approaches, is the part of this piece worth reading in full.
The bet here is that the next wave of AI is won by whoever solves data efficiency, not whoever builds the biggest cluster. Flapping Airplanes is making multiple directional bets toward biologically-inspired model architectures to get there. Whether a team this young can execute on a 5 to 10 year research agenda against labs with orders of magnitude more capital is the open question Cahn does not fully answer.
[READ ORIGINAL →]