Hi, I’m Greg. I was recently laid off from Block, and one of the biggest ideas I left with came from what I saw there: AI adoption across engineering teams was incredibly uneven, and the people getting the most out of tools like Codex and Claude Code were often the hardest to learn from. I sat through a lot of demos, calls, recordings, and screen shares where engineers shared impressive AI workflows. But most of the time, you could only see the outcome. You couldn’t really inspect what they typed, how they guided the agent, what they tried, what failed, or why they made certain decisions. That gap is what led me to build athrd. athrd automatically attaches AI coding sessions to commits and surfaces them in pull requests, so reviewers can understand not just what changed, but why it changed. Instead of reviewing a diff in isolation, teams can see the context, tradeoffs, and decision trail behind the code. It also turns strong AI workflows into shared team knowledge. The best AI-native engineers stop being black boxes, and their process becomes something the rest of the team can actually learn from. athrd fits into the workflow teams already use and helps make AI-assisted development more reviewable, teachable, and trustworthy. If you’re using Codex or Claude Code, I’d love your feedback. ps: Codex help me write this and you can see it here https://www.athrd.com/threads/64...

ATHRD

AI traceability for commits and pull requests.

AI traceability for commits and pull requests.