August 18, 2026Assessments
Grading from each candidate's assigned questions
Candidates are now graded against the questions they were given, even if the assessment was edited after they started. The candidates table and the candidate's report now show the same score.
What changed
- Candidates are now graded against the questions they were given, even if the assessment was edited after they started. The candidates table and the candidate's report now show the same score.
- Files created in a question's sandbox, including ones Claude Code or Codex write, are now saved with the candidate's submission. When the AI agent runs a command in the sandbox, the terminal now shows its output even if the tool also printed warnings or progress.
- Recruiting Copilot now keeps each question's AI level when a request mixes levels, such as one Rena question and three Agentic ones. It also groups a reply's web sources into one list.