Research
How the best AI companies get built.
We sit inside the VC network and recruit the people frontier teams are made of. This is what we see: deep team teardowns and field notes on the talent market underneath the headlines.
Lila Sciences, Explained
Periodic hired OpenAI. Lila hired five university labs. Ninety-eight of 535 staff read across a $550M AI-for-science company — professors as chiefs, their benches following them in, and a hard turn toward the factory in 2026.
Simile, Explained
A Stanford paper turned into a company — then sold by the field force of the category it replaces. Fifty-nine people, $300M at a $2B valuation, and 26 who sell or deploy against 23 who build.
Periodic Labs, Explained
A frontier-AI research crew, a robotic chemistry lab and $605M — aimed at one material that doesn’t exist yet. Seventy-one people hired in the order the bets required: bits in 2025, atoms in 2026.
The Barbell Company — Moonlake AI
Three olympiad champions and a 25-year NVIDIA Distinguished Engineer in the same 30-person startup — and just 4 of the 31 profiles we read are anywhere near the middle of their career.
Old Hands, New Physics — Unconventional AI
How a $475M seed welded two alumni networks that had never met — the founder's own acquisition chain and the analog bench that built Wi-Fi — into a 43-person physics-chip team.
I screened 237 resumes from Stanford and Berkeley. Here's what still separates the best.
When everyone looks strong on paper, the resume stops being a signal. Three patterns that still tell the real top talent apart.
Xemini Assemble — Elorian AI
How the researchers behind Gemini's data and Apple's foundation models reunited to chase native visual reasoning.
Anatomy of a $643M Team
How a ~20-person MIT spinout built the full inference stack and sold to Nebius for $643M in ten months.