Free practice for AI allowed coding interviews

Practise the coding round where AI is allowed

Some interviews now let you use an AI assistant, and some ask for your AI transcript. Open Book gives you three realistic 60 minute rounds to run with your own agent, a rubric built from what interviewers say they grade, and a tool that counts what happened in your session without uploading it.

3
free practice rounds, with checks
18
company policies, each with sources
12
answers to common questions
0
bytes of your transcript uploaded
Start a practice round See which companies allow AI

The format

What an AI allowed round is, and who runs one

In an AI allowed round you solve a realistic problem with an AI assistant beside you, and the interviewer watches how you use it: what you ask for, what you check, what you reject. It is still rare. These employers run one, each with its source on the company page:

  • Canva: requires AI in its AI Assisted Coding interview, in Canva's own words (June 2025).
  • Meta: one of the two onsite coding rounds has been AI enabled since October 2025, according to prep sites.
  • Shopify: encourages AI in remote interviews, according to its head of engineering (July 2025).
  • Rippling: AI is optional, with a different rubric when you use it, according to a prep site.
  • Google: a pilot from the second half of 2026 for some junior and mid level US roles, according to press reports.
  • Rootly: expects AI on its take home and asks for the AI transcripts, in Rootly's own words (February 2026).

Most of the largest employers have changed far less. In a 2025 survey by interviewing.io, none of 52 interviewers at big tech companies said their company had dropped algorithm questions, while 67% of startup respondents said AI had meaningfully changed their process. Amazon and Anthropic ban AI in interviews unless they say otherwise. See all 18 company policies.

Sources: interviewing.io's 2025 interviewer survey for the survey figures; each company page lists its own sources. All checked 18 October 2026.

Free practice

Three practice rounds, 60 minutes each

Each round is a small, realistic codebase in Python with the standard library only, a brief that leaves some decisions to you, tests that fail until the work is done, and extra checks to run when you finish, including one production trap the brief does not mention.

How to practise

Run it like the real thing, then review how you worked

  1. Use your own agent

    Claude Code, Cursor, Codex or any other assistant you would use at work. If your interview gives you a chat only assistant, as Meta's reportedly does, practise that way too.

  2. Set a 60 minute timer

    Read the brief and the code before you prompt. Write the decisions you make in a notes file. Do not open the checks folder.

  3. Run the checks

    When the timer stops, run the extra checks. They test what a reviewer would test before the change ships.

  4. Review the session

    Paste the prompt from REVIEW.md into your agent so it reads your transcript against the rubric and quotes your own moments back to you. Drop the transcript on the stats tool for counts.

The rubric

Five things to look for in your own transcript

Built from what employers say they read, such as Rootly's four review points and Meta's reported competencies, and from a study of 19 Microsoft developers working with an agent. It is a practice rubric, not a validated score: no study has validated one.

  1. Breaking the problem down. Turning a loose brief into a plan you own, then handing the AI one well defined step at a time.
  2. Context given to the AI. Supplying the facts the AI cannot guess: the constraint in a config file, the rule in a document, how the system is deployed, the helper that already exists.
  3. Verifying output. Checking that what the AI produced is right before relying on it.
  4. Correcting the AI. Noticing when the AI goes wrong, stopping it, and steering it back precisely, or doing the small fix by hand.
  5. Tradeoff reasoning. Naming the options, choosing one for a reason tied to this system, and saying what was given up.

Read the full rubric

Transcript stats

Counts from your session, read in your browser

Drop a Claude Code session file or a Codex CLI rollout file on the stats page and it counts what happened: prompts, tool calls by type, edits you rejected, test runs, plan mode, subagents and session length. Nothing is uploaded. These are counts, not a score. The worked example, a made up session, shows:

6prompts typed
18tool calls
2tool calls rejected or interrupted
4test or check commands
47minutes

Count your own session

Stay in touch

New rounds by email, or a reviewed plan

New rounds by email

Free. One email when a new practice round is published, and nothing else.

We store your email address and the choices above, with the date. Privacy, including how to remove it.

Reviewed rounds, $19 a month

$19 a month

Two practice rounds a month reviewed by a senior engineer within 48 hours, with the moments in your transcript that cost or earned you points quoted back.

Not open yet. Nothing is charged and no card is asked for: this only records that you would pay $19 a month, so we know whether to build it.

We store your email address and the choices above, with the date. Privacy, including how to remove it.

Answers

Questions people ask about AI in interviews