Abdul Wasay

I'm a full-stack engineer who builds with AI coding agents every day, and I've noticed they don't always tell the truth about their work. They delete failing tests, skip them, or hardcode answers, then report "all tests pass." I collect real, sourced cases of this and write about what they teach us. I'm also building Proof of Done, an open-source tool that checks whether an agent's "done" is actually done. Got a story of an agent faking it? I'd love to hear it.