Y Combinator

Seed accelerator whose stated theses and Requests for Startups tend to precede broad market
attention on AI by several quarters.

What they do

Twice-yearly accelerator batches with seed investment and a large alumni network. Alongside the
investment business, YC publishes research reading (Paper Club), founder interviews, the
Lightcone discussion series, and Requests for Startups — an explicit list of problems they
want funded.

Why they matter in AI

Requests for Startups and the batch composition function as a leading indicator: what YC funds in
one cycle is often a mainstream product category two or three cycles later.

The clearest example is the reasoning-over-scaling call. In late 2024 YC published the argument
that the next AI breakthroughs would come from reasoning rather than parameter scaling — well ahead
of the reasoning-model wave that followed. Worth weighting their stated theses accordingly.

Coverage worth separating

Three distinct registers, with different reliability:

RegisterReliability
Paper Club — structured reading of primary researchClosest to primary; low spin
Policy — e.g. White House AI strategy discussionsUseful access, treat as sourced reporting
Founder interviewsAdvocacy — the speakers are pitching