hello millie

RESEARCH LAB · TAILORED BENCHMARKS · NO REGRESSIONS

The frontier,
measured.

A research lab that delivers the highest-performing models into your business processes. We benchmark on your workflows, not a public leaderboard. When models change, your results don't slip.

Request an invitation VIEW THE BENCHMARK →

THE DIFFERENCE

Your processes are the benchmark.

  1. 01

    Measured

    We don't score models on trivia. Your workflows are the test suite: the invoices, the handoffs, the edge cases that make your business yours.

  2. 02

    Proven

    Every night, the benchmark runs against your processes. When models update, the change ships only if your scores hold or improve. Updates make things better, never worse.

  3. 03

    Delivered

    Research only counts in production. We ship the winning model into your processes, then keep measuring it there.

CONTENTS

  1. 01How we measure
  2. 02The dumb zone
  3. 03How agents remember
  4. 04Three kinds of RAG
  5. 05The frontier, priced
  6. 06Big model, small model

FIELD NOTES · hello millie · 2026

Work with the lab.

Request an invitation