This month's theme: a smoother path for test-takers and faster authoring for your team. October brought a more streamlined assessment experience, Cosmo support for nearly every question type, and practice questions that help test-takers prepare for agentic assessments.
Highlights
- A smoother assessment experience for test-takers. Test-takers now go straight into their assessment without a separate home page, practice questions align with the real assessment more closely, and AI interviews now show the assessment title, a clear submission status, and an easy way to end the session. If your team runs internal test-takes, you'll notice the updated flow.
- Cosmo creates almost any question type as you build an assessment. Cosmo can now draft quiz, writing, and conversation questions on the fly while you create an assessment, and can help with coding questions too (available to users who can create questions for their organization).
- Practice for agentic assessments. When a test-taker's upcoming assessment includes an agentic question, their practice set now includes an agentic practice question too, so they can get comfortable with the format before the real thing.
Assessments & Hiring
- Streamlined test-taker experience. No more home page step, practice that mirrors the assessment, and a simpler AI interview screen. Test-takers can also tell whether a conversation answer has been submitted, and know their highest-scoring attempt is the one that counts.
- Agentic practice questions. These appear only in the practice set tied to an assessment that includes an agentic question.
- Fewer false suspicious-paste flags. Text pasted from the Cosmo chat, or code copied with the "Copy code" button in a question description, no longer counts as a suspicious paste, so Suspicion Score reports are easier to review.
AI Interviews & Conversation Questions
- Set attempt limits on conversation questions. You can now choose how many attempts test-takers get on a conversation question in AI interviews and assessments. Test-takers can see how many attempts remain, and the default stays unlimited.
Questions, IDE & Simulation
- Cosmo authoring for more question types. Cosmo is most helpful for non-coding questions. Complex filesystem questions with multiple environments are still best built in the question editor.
- Hide the debug panel per question. Question authors can now hide the debug panel on individual questions in the new CodeSignal IDE.
- Prepared interview workspaces stay put. Interviewers can set up a Free Filesystem question ahead of time and step away. Their files are kept until the interview starts.
- Custom grading for custom simulations. Custom simulations can now score test-takers using their own grading logic.
- Refinements to custom simulations and interviews. Custom simulations now display their own name throughout the experience, progressive filesystem questions launched from an interview template move through levels just as they do when opened directly, interview environment setup is smoother, and file names that include brackets save exactly as typed.
Integrations & API
- Greenhouse integration pauses safely when credentials stop working. After repeated authentication failures or an expired authorization, the Greenhouse Harvest integration pauses automatically and admins get a banner and an email explaining why. Saving valid credentials turns it back on.
- Consistent status for rejected proctored assessments. A rejected proctored assessment now stays "rejected" in Greenhouse, SmartRecruiters, Jobvite, iCIMS, and Oracle, so recruiter automation and test-taker follow-up stay on track.
- Clearer skill-area data in the API. The Coding Score skill-area breakdown now labels every level of a progressive question, including the first. Test-taker scores are unchanged.
Learn & Content
- New CodeSignal IDE rolling out in Learn. The new coding environment is live on several Learn course paths, with tabs, custom settings, a better terminal, and more stable sessions. More courses will follow.
- New assessments, including LLM Prompt Evaluation. The LLM Prompt Evaluation Assessment is one of several new assessments added to our library this month. It measures a test-taker's ability to write complex LLM prompts for a real domain need and back them up with evaluations that show the prompts work consistently.