This illustrative packet shows what a hiring team inspects after an async DevEval screen: executable results, AI judgment, replay context, integrity signals, candidate-safe shareback, and a human next step.
DevEval does not ask reviewers to trust a single score. Each signal links back to the work, the failure mode, and the reviewer action.
Deterministic results
The candidate solved the core task and exposed the failing edge cases that should drive follow-up.
Reviewing AI output
The candidate caught three of four issues in the AI-generated code and explained the risk clearly.
AI-era workflow
Prompts asked for constraints, tests, and failure modes instead of asking the model to write code blindly.
Reviewable events
The report shows session context without turning surveillance into an automatic hiring decision.
After a final hiring decision, teams can deliberately release a private, score-free Candidate Growth Report while keeping rubric evidence and selection reasoning inside the hiring workspace.
Private Growth Report
Avery gets practical coaching on what went well, what to improve, and how to practice the same skill. Scores, rankings, hiring outcomes, integrity signals, and reviewer notes stay private to the hiring team.
Read instructions and opened starter tests.
Merged arrays correctly for sorted positive inputs.
Flagged race condition and missing null guard.
Duplicate key eviction and empty fallback response remained.
Added time complexity and named the unresolved risk.
A reviewer can advance the candidate to focused live validation instead of repeating the async screen.
Decision owner
Human reviewer
Score use
Evidence summary
Risk flag
Boundary case