CodeVerdict is an AI-powered platform that reviews developer take-home coding assignments in under a minute. You drop the assignment brief and a GitHub URL (or zip); CodeVerdict maps every requirement to the code, detects AI-written shortcuts, runs the project in a sandbox, and produces a scored report with tailored interview questions.

How long does it take to grade a take-home assignment?

Most assignments are reviewed in 60 seconds or less. The agent reads the brief, clones the repo, runs setup / tests / execution in an E2B sandbox, and writes the report in parallel. Larger repos with heavy dependencies may take up to two minutes.

What can I submit — GitHub repos, zip files, or both?

Both. You can paste one or more GitHub repository URLs (public or private with a GitHub token) or upload a zip file of the candidate's solution. CodeVerdict handles each the same way: clone, install, run, evaluate.

Do I need an account to try it?

Yes — a free CodeVerdict account is required. Sign up takes about 10 seconds via Google one-click. Accounts let your assessments sync across devices and your team see the same dashboard.

How does CodeVerdict detect AI-written code?

It combines three signals: token-level perplexity (LLM-written code has unusually uniform probability distributions), naming and structural entropy (AI tends to use overly consistent patterns), and commit-history analysis (sudden large commits with no incremental work). The output is a calibrated 0–100 AI score, not a black-box verdict.

Is candidate code safe to run? Where does it execute?

Yes. Every repository runs inside an isolated E2B sandbox — a single-use virtual machine that is destroyed immediately after the report is generated. The candidate's code never touches your machine or CodeVerdict's production servers.

Which programming languages and stacks are supported?

CodeVerdict supports Node / TypeScript, Python, Go, static front-ends, and any Docker-based project out of the box. The agent detects the stack automatically from project hints (package.json, requirements.txt, go.mod, Dockerfile, etc.) and picks the right setup path.

Can I customise how candidates are scored?

Yes. The displayed score is computed client-side from your weightings — requirements met, code quality, test coverage, security, AI-written code. Adjust the weights in Settings and the verdict (Strong hire / Hire / No hire) updates instantly across every submission.

We tried using AI to detect AI-generated code. It didn't work.

We tried using AI to detect AI-generated code.

It didn't work.

Here's what it actually told us: → "This code is too perfect for a normal person" → "The folder structure is unusually clean" → "A typical developer wouldn't write this well"

That's not detection. That's bias. A good developer would get flagged. A clever cheater wouldn't.

So we scrapped it completely.

We went back to basics.

We started reading hundreds of AI-generated submissions manually. Patterns showed up fast.

→ Overly formal comments explaining obvious things → Massive block comments on every function → Repetitive code a human would just abstract away → Params style blocks on every single function → Comment style that reads like documentation for strangers

Humans comment like they're leaving notes for themselves. AI comments like it's writing a textbook.

The new detection layer

We rebuilt the detection layer from scratch.

No LLM call. Just regex pattern matching across the entire codebase. When something unusual is detected we surface the complete code block to the reviewer. Not a verdict. Not a score. Just "look at this, you decide."

The missing signal: Git commits

Then we added one more signal: Git commits.

AI-assisted commits are huge. "Implemented full authentication system with JWT, refresh tokens, middleware and role-based access control" in one commit.

Human commits are small and honest:

"fix login bug"
"forgot to add env"
"wip"

The result

→ More accurate than our LLM approach → One less API call on every single assessment → Faster, cheaper, and more honest

Sometimes the smartest solution is the simplest one.