All experiments
Labs · Research

Learning from reviews

What you edit, reject or send back should make the next run better.

Why it matters

The problem we're chasing

Every override is a signal. A system that learns from what humans correct — without training on your private data — could calibrate its own proposals over time, and eventually tell you how reliable they are.

What we're exploring

The questions, not the recipe

We publish the direction we're working in — not how we build it. The mechanism is what makes TaskForce different, so it stays in the lab until it ships.

  • H1 Turning overrides into signal
  • H2 Calibrating proposals without training on your data
  • H3 Measuring whether a decision was actually right

Where it stands. This is the Prediction & calibration direction — Planned, not shipped. It's the honest edge of what we're building.

Follow where this goes

The roadmap is public. Hold us to it — and start with what ships today.