PromptProven · pilot

Hire engineers who can direct and check an AI assistant.

PromptProven is a take-home assessment for teams hiring software engineers. Candidates use a built-in AI coding assistant to fix, change or extend a real codebase, and their work is graded against tests they never see and on how they worked.

How an assessment works

  1. 01You choose tasks from the catalog and invite candidates by email.
  2. 02Candidates work in a hosted workspace in the browser: the task’s source, an editor, a terminal and a built-in AI assistant. The session is timed and starts with a short written plan.
  3. 03The submitted change is built and run against hidden tests in an isolated container, and the recorded session is scored for how the candidate worked with the assistant.
  4. 04You review every candidate’s results, task by task, in one place.

What we are building with pilot teams

Passing tests shows that a fix works. It does not show whether the engineer understood the problem, steered the assistant, or caught its mistakes. That is why each session is time-boxed and recorded, and half of the score is for that process. We are validating the process score with a small number of teams before calling it a product.