Claude Code & Codex Usage Trading Cards by Rudel

For folks self-hosting — what's the resource floor? Single node or does it expect a small cluster? And how does upgrade work once a real team is depending on it — drop-in replace the binary, or is there a migration step?

very cool product, congrats Evren!

hey all!we built rudel because we were using claude code / codex every day, but had no idea what was actually happening across sessions.which sessions worked?which ones got abandoned?where were tokens going?were we getting better, or just spending more?we launched a month ago and had great reception.so we tried something more fun: spotify wrapped x fifa ultimate team cards for claude code and codex.upload your sessions, and rudel gives you your ai coder card.some examples:roadrunner: fast, frequent, high-output sessionstourist: light usage, lots of starts, low commitmentcompany card: high intensity, high spend, meh outputmaniac: broad, consistent, intense usage across reposadhd brain (me): lots of repos, mid to low throughput acrossand 4 more...the classifier runs on derived metadata like duration, token counts, model mix, repo count, and commit signals.free and open source.would love feedback on the cards, the archetypes, and what else you’d want to understand about your claude code / codex usage.cheers!