Neolabs today face an impossible mandate: deliver both a research breakthrough and a venture-scale business, even though each is a 1 in 100 outcome. Combined, the odds drop to 1 in 10,000. The author argues for a better model borrowed from pharma.
Frontier labs like Open AI, Anthropic, and Google Deep Mind are locked in a race to ship proven improvements—data pipelines, infrastructure, compute—while researchers chase "the next big thing." This tension has spawned neolabs, startups funded heavily before having a product, where teams split between building a company and pursuing AGI.
The author proposes a radical shift: let neolabs focus solely on groundbreaking research, expect near-zero revenue, and accept acquihires if they succeed. This mirrors pharma's decades-old division of labor, where small biotechs handle risky discovery and big companies scale production and distribution.
While this works for science, it fails VCs, who need fund-returning companies rather than acquisitions, especially at today's multibillion valuations.
Source: Hacker News · Summarized by HeadlinesBriefing