Peter Steinberger (steipete), OpenClaw creator and OpenAI employee, burned $1.3M in raw API costs over 30 days, visualized via a CodexBar update.
Key Takeaways
CodexBar now renders API cost breakdowns visually; Steinberger used it to surface the $1.3M figure.
Spend implies roughly 600B tokens in 30 days, averaging ~$19K/day at retail API rates.
Steinberger works at OpenAI, making the dollar figure a rack-rate accounting artifact rather than out-of-pocket cost.
OpenClaw relies heavily on the best available models; capability degrades sharply on smaller or cheaper alternatives.
The spend is partly driven by automated agent pipelines and repo-policing automations, not just interactive coding.
Hacker News Comment Review
Commenters broadly flagged the headline as misleading: Steinberger is an OpenAI employee with unlimited access, so $1.3M reflects list-price token math, not actual spend or subsidy cost.
A recurring thread debated productivity ROI: his GitHub shows 8,826 commits across 94 repos in May 2026, but critics argue the output reflects marketing velocity more than engineering depth.
Practical builders noted that retail users hit spend limits fast – one commenter spent $40 just wiring Google auth tokens in OpenClaw on a VPS, and it broke within a week, underscoring the gap between Steinberger’s access and normal operator economics.
Notable Comments
@thomasahle: At 60 Codex subscription accounts ($200/mo each), the same token volume costs ~$12K/month – a 100x reduction versus retail API pricing.
@yodakohl: Points to the public output – 8,826 commits in 94 repos in May 2026 alone – as concrete evidence of agent throughput at this scale.