This page runs CodeMind's own public @codemind/ai pipeline — requirements extraction, per-requirement verdicts with resume quotes, and code-computed scoring — unmodified, against a real model. Pick a pair and run it.
Step 2 — Verdict per requirement (checked against the resume)
Verdicts stream in as each one lands — the UX from suggestion #1 in my review.
0
overall ATS score (computed in code, not by the model)
How this was built: the public repo codemind-hq/codemind-ai (commit 927fc00), zero library changes — only a small client adapter so the library's own gateway talks to a hosted model (GPT-5.6 Luna via ParsPack AI Studio). The runs shown are real recorded runs of that pipeline (Oct 22, 2026), each costing ≈ $0.003 in metered model usage. Quotes are verified against the resume text by the library's own checks — a quote the resume doesn't contain drops the verdict.