Vetting · 12 of 2,000

12 of 2,000 make it through.

We contact 2,000+ senior AI engineers each month. We accept 12. Every engineer we accept survives a 5-stage process. This page shows the high-level shape. The full rubric is ours, shared under NDA.

✦ The funnelEach month
  • Contacted
    2,000+
  • Screened
    ~250
  • Advanced
    ~30
  • Accepted
    12

Five stages.

Headlines only. Each stage has its own rubric, scoring, and pass/fail thresholds. We don’t publish those. Three things matter most: production taste, AI fluency and behavior under ambiguity.

  1. 01

    Initial screen

    Sourced from active AI builder communities, GitHub contributors, and direct outreach. Filtered on years of senior production experience and shipped AI surface area.

  2. 02

    Technical assessment

    Production code review. Real systems they shipped. Not LeetCode. We’re looking for taste, defensive coding, and sound tradeoffs, not algorithm puzzles.

  3. 03

    EQ + behavioral

    How they communicate, push back on PRs, ask questions when things are ambiguous, and behave when they don’t know the answer. Engineers without strong EQ break partnerships quickly.

  4. 04

    Paired AI challenge

    A live, scoped problem we pair on with the candidate. Real Cursor + Claude session. We watch how they think with AI tools, not just whether they can use them.

  5. 05

    Final filter: Jess

    Jess Mah (Data Scientist · UC Berkeley CS at 19) runs the final technical conversation herself. Every accepted engineer clears her bar. No exceptions. This stage also checks references, compensation alignment, and cultural fit with the teams we build with.

✦ A note on transparency

We don’t publish our full rubric, for the same reason any serious selection process keeps its scoring private: it’s the IP. What we will share, on the first call, under NDA: per-stage benchmarks, sample assessments, and the exact bar a senior AI engineer has to clear to work on one of our partners’ projects.

Want the full rubric?

It’s not on the public site by design. We share the full assessment criteria, sample evaluations, and per-stage benchmarks on the first call, under NDA.

Every proposal is read by one of our founders.

Let’s talk →Send a written brief →