Reviewing submissions

You review every candidate from the source assessment. Reviewers never see candidate clones in the workspace switcher; open the source and work from its sidebar. As candidates submit, their submissions appear there; opening one gives you the submitted files, the session recording, and the AI usage transcript. File review is read-only: nothing you do changes what the candidate submitted. Scores are autograded onto the clone after submit; you can edit those chips without changing the files.

Where the source lives

Find the source in the workspace switcher:

• Assessments - at least one accepted clone is not started, in progress, or submitted but not yet graded.
• Past assessments - muted, same pattern as Past interviews. The source moves here once every accepted clone is submitted and graded (autograde auto-apply or a human score). Submit alone is not enough.

Opening a finished source shows a Past assessment badge in the header. Finished live interviews use Past interviews and a Past interview badge the same way.

An invite that was never accepted does not keep the source under Assessments. Candidates see their own clone under Past assessments as soon as they submit, even before a grade lands.

You review from the source sidebar either way. See Take-home assessments overview.

Browsing the submitted files

Open a submission to browse its files in a read-only editor that looks and behaves like the normal file tree. You see exactly what the candidate left at submission time.

• Markdown files (a README.md, a DESIGN.md) open in a rendered preview by default, with a Source / Preview toggle. A candidate’s written explanation shows as formatted prose, not raw markup.
• The view is a snapshot. Use the sidebar refresh to re-list files if you are looking at an in-progress candidate; re-open a file to re-read its latest content.

Replaying the recording

If recording was enabled for the assessment, each candidate’s session is captured automatically and can be replayed once they submit. The recording shows the edits, terminal output, and test runs over the course of the session, so you can watch how the solution came together rather than guessing from the final state.

Recordings are listed on the source assessment and tagged with the candidate, so you replay the right session for the right person. See Recording & replay for how recording works in general.

Reading the AI transcript

If the AI assistant was enabled, open the AI transcript for a submission to see the candidate’s exchange with it: every prompt they typed and every reply, in order, with each candidate identified.

• Machine-context noise (system reminders, tool results) is stripped out and tucked behind collapsible sections, so you read the actual conversation.
• A Raw toggle shows the unmodified stored content if you want to audit exactly what was sent and received.

The transcript is the basis for grading how the candidate used AI, covered next.

What stays hidden

Everything reviewer-side stays reviewer-side. Candidates never see the rubric, scores, recordings, transcripts, or one another. See Sharing & roles.

Next: Grading.