Code review, debugging AI-generated code, and evaluating model outputs across programming languages — at the highest live rate ceiling on the platform.
Tier: Coding (the highest live rate ceiling on the platform)
Coding subcontractors review code that AI models generate — spotting bugs, evaluating idiomatic correctness, checking edge cases, and judging whether the output would actually run in production. The work spans Python, JavaScript, TypeScript, Go, Rust, Java, C++, SQL, and the most common day-to-day languages.
Coding is the highest-paying specialty on Jobpeak because the work demands the sharpest disposition: catching bugs that look plausible. Models confidently produce code that compiles but fails at runtime, code that handles the happy path but breaks on the edge case, and code that uses APIs that don't exist. Your job is to flag those outputs at the speed the model produces them.
The Cognition Test for Coding draws from a 150-question bank covering language fundamentals, debugging, code review, and AI-output evaluation across the most common programming languages.
10 easy / 20 medium / 10 hard in the domain section. Behavioural section unweighted by difficulty.
Spot-the-bug exercises in Python, JavaScript, and SQL; idiomatic-vs-non-idiomatic judgements; code-review tasks where the brief is intentionally ambiguous; identifying hallucinated APIs and made-up library functions.
Practical fluency over theoretical knowledge. We don't test you on the most obscure language features — we test you on the everyday judgements that catch model errors at scale.
Qualified Coding subcontractors are matched to live AI review work within seven days. Rate ceiling is up to $100/hr; the matching system surfaces work as it becomes available.
What you need to know about this specialty before you sit the test.