Home

Training & evaluation

A clear brief for every capability.

Explore six project directions. Each brief explains the task, expected deliverables, and the evidence needed to evaluate the result.

Training & Evaluation

Built Around Real Work

Six ways to turn a capability gap into a concrete task. Explore illustrative briefs, delivery formats, and evaluation methods.

Coding environments

Code that survives the checks.

Explore repository tasks with a runnable starting state, an explicit output contract, and deterministic checks.

Explore the approach

Agent workflows

A workspace. A task. A trace.

Explore longer tasks that require an agent to reconcile documents, manage state, and deliver a usable artifact.

Explore the approach

Expert reasoning

Make the reasoning inspectable.

Explore context-dependent tasks where each conclusion must follow the definitions and evidence supplied in the brief.

Explore the approach

Search & report evaluation

Evidence behind every answer.

Explore open-ended research tasks with explicit standards for source quality, useful synthesis, and fit to the user's needs.

Explore the approach

Domain science

Measure judgment under constraints.

Explore scientific evaluation that makes the trade-offs, missing evidence, and next experiment part of the task.

Explore the approach

Multimodal & 3D

Evaluate the editable result.

Explore creation tasks that preserve the brief, execution process, visual checks, and an editable final project.

Explore the approach

Start a conversation

What should your AI do better?

Talk to our team