Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Anthropic's AI biolab finds 'CRISPR-like' DNA in viruses. What's next? (nature.com)
    —discuss
  2. Recurse – Develop and deploy specialist agents faster (recurse.run)
    1comments
  3. Skill-Guided Mining and Compilation of LLM Agent Traces (arxiv.org)
    —discuss
  4. AI Companies Are in a Race Against Time (econjared.substack.com)
    —discuss
  5. Locked Down Passkey and Keychain Backups (mjtsai.com)
    —discuss
  6. Show HN: OnionGuard – IP-agnostic, zero-JS DDoS defense in Go for Tor services (github.com/ihatemyfcklife)
    —discuss
  7. Rank Anything: convert your income to League of Legends rank (github.com/szge)
    2comments
  8. AI Workers' Inquiry 2026 (techworkersinquiry.org)
    —discuss
  9. Show HN: VNetMap – A zero-knowledge E2E encrypted network topology mapper
    —discuss
  10. AI often makes writing worse. (Even if it makes it better) (economist.com)
    —discuss
  11. Neuroergonomics (2011) (youtube.com)
    1comments
  12. Show HN: Jev.Store – A dedicated directory of useful things made on Jev (jev.store)
    —discuss
  13. Kiro adds jev to decide turn in preview (kiro.dev)
    —discuss
  14. Initial-scale no longer needed (quirksmode.org)
    1comments
  15. WBHM team expected to use "Gulf of America" in place of "Gulf of Mexico" (uab.edu)
    —discuss
  16. Is z.ai shipping a different ZCode client than the source?
    —discuss
  17. Mission-Bound Authorization for OAuth 2.0 (ietf.org)
    —discuss
  18. How to copy URLs of multiple Firefox tabs without installing an extension (autodidacts.io)
    —discuss
  19. What Even Is an OS Now? (sockpuppet.org)
    —discuss
  20. AI and human tutoring yield equivalent GRE learning gains (arxiv.org)
    1comments
  21. IBM When is Watson going to regain the crown via television?
    2comments
  22. It's not hypothetical: the dangers of AI are here (theguardian.com)
    —discuss
  23. Minimal Assumptions for Societal Resilience (chillphysicsenjoyer.substack.com)
    —discuss
  24. Show HN: Farm.js – A full-stack framework with an AOT compiler for React (farmjs.dev)
    —discuss
  25. Carnegie: China Passes US as Top AI Talent Hub, 40.6% to 34.2% (aiweekly.co)
    1comments
  26. Show HN: NerfWatch() – Track AI degradation with daily tests and community votes (nerfwatch.lol)
    —discuss
  27. Show HN: BuyerCue – Find potential customer by paid-placement evidence (buyercue.io)
    1comments
  28. Where's the Beef? [video] (youtube.com)
    —discuss
  29. Kafgres: Embedding a Kafka Broker into Postgres (rynr.dev)
    —discuss
  30. The Animated Elliptic Curve (xargs.org)
    —discuss

FYI: OpenAI "spend limits" aren't always limits

1 pointsby 1h ago
0 comments
My OpenAI API key got hacked this morning, and the bot started using my key for something in Chinese.

"No worries," I thought, "I have a $30 spend limit setup." When I came back to my desk an hour later, I had $285 in unauthorized charges.

As it turns out, OpenAI will let you create a spend limit _without_ checking "enforce spend limit." If you don't check enforce, it's meaningless. So even though my account showed that I had a $30 spend limit, it still let the hacker blow right past it by hundreds of dollars.

I've submitted a dispute, but wanted to let other people know so it doesn't bite you, too. You can check your spend limits at https://platform.openai.com/settings/organization/limits

A quiet thread, for now.Start the conversation on HN ↗