Claude ranks third in Parameter Golf
OpenAI's Parameter Golf challenge asks participants to fit a pretrained language model into a 16 MB artifact and train it in under 10 minutes on 8xH100s. The Claude contributor ranking mostly reflects the repo's PR approval flow, not OpenAI's internal tooling.
The viral angle is mostly optics: contributor charts are a weak proxy for who built what when a repo accepts outside PRs and merges approved results onto a public leaderboard. OpenAI is using the challenge as a crowd-powered idea sieve and recruiting funnel, with participants forking the repo, submitting code, logs, score, and a write-up before approval, and the repo also offering Runpod compute credits and a hiring hook. HN commenters note that Claude's attribution is tiny and comes from accepted submissions, so the third-place stat is easy to overread.
DISCOVERED
17d ago
2026-03-25
PUBLISHED
17d ago
2026-03-25
RELEVANCE
AUTHOR
mirzap