Counterweight Labs

01

Realistic tasks

Train on security work that requires planning, tool use, and actions across a live environment.

02

Verifiable rewards

Use the resulting system state to determine whether an objective was reached, a threat removed, or a service restored.

03

Repeatable experiments

Start from controlled conditions, vary the scenario, and compare behavior across training runs.

The engagement

What your team gets

01

An environment package

The systems, task definitions, starting conditions, and dependencies required for the agreed capability.

02

Grading and reward logic

Checks that turn verified outcomes into a training signal, with failure cases and limitations documented.

03

Reset and scenario design

Defined initialization and reset behavior, plus scenario variations suited to your training objective.

04

Integration and handoff

An agreed interface to your training stack, documentation, and support scope for adopting the environment.

The training loop

How the training loop works

  1. 01

    Agent

    A model acts through the agreed tools.

  2. 02

    Environment

    Actions change the state of a real system.

  3. 03

    Grader

    Independent checks verify the outcome.

  4. 04

    Reward

    The training stack receives the result.

Working with us

From scoping
to delivery.

  1. 01

    Define the capability

    Identify the behaviors you want to develop, the agent’s tools, and the systems it needs to interact with.

  2. 02

    Scope the environment

    Agree on task coverage, reward design, deployment requirements, licensing, and integration with your training stack.

  3. 03

    Integrate and iterate

    Validate the environment and reward behavior, then use training evidence to guide scenario development.

Common questions

Who are these environments for?

Frontier model companies and teams developing cyber-capable agents. Engagements focus on the model, training stack, and capability objectives you need to support.

Can you build a custom environment?

Yes. We can scope an environment around an offensive or defensive capability, with explicit tools, starting access, objectives, and reward criteria.

Does a license include the benchmark test set?

No. Training and evaluation can share infrastructure and grading interfaces, but held-out benchmark tasks, seeds, answers, and evaluation traces stay separate.

Start a conversation

What should your next model be able to do?

Tell us what you are measuring or training, and the systems your team needs to work with.

Discuss environments