In this role, you'll reimplement a software package end-to-end without original source code or AI tools, to benchmark human performance.
Ensure your code produces output matching originals exactly on all tests, including held-out tests.
Work across diverse computing domains: Unix utilities, data serialization, bioinformatics, interpreters, static analysis, cryptography, and compression.
Write all code without AI assistance based on precise written specifications.
Contribute data to measure human capabilities against advanced AI systems.