Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Why open source always wins(pgdog.dev ↗)
    discuss
  2. Show HN: I made a game about awkward conversations using Jev(playuhoh.com ↗)
    1comments
  3. New(pirateface.co ↗)
    discuss
  4. Microsoft's new AI 'code of conduct' tells models not to hack systems(techcrunch.com ↗)
    1comments
  5. Just uploaded **Chameleon v0.1.1 Beta(mrjaki.github.io ↗)
    discuss
  6. M5 Ultra Mac Studio Review: The Dream Mac for Local AI Agents(macstories.net ↗)
    discuss
  7. An Undercover Google Analyst Infiltrated a Notorious Supply-Chain Hacking Gang(wired.com ↗)
    discuss
  8. Canonical announces Zephyr 26.04 LTS(canonical.com ↗)
    1comments
  9. Show HN: Run Jev-style models locally on Mac with 0.74 GB RAM(github.com/afshinm ↗)
    discuss
  10. OpenAI President Greg Brockman on Doing Business in the Wake of Hugging Face [video](youtube.com ↗)
    discuss
  11. ROG Xbox Ally X20 / Tokyo Games Show 2026: Improving, but No SteamOS Planned(boilingsteam.com ↗)
    discuss
  12. Shopify CEO says employees' 'slop grenades' are making more work for everyone(businessinsider.com ↗)
    discuss
  13. EU KIDS Act Won't Keep the Internet Accountable and Trustworthy(eff.org ↗)
    discuss
  14. Step 5 Preview: StepFun's frontier model with 1M context and video input(stepfun.ai ↗)
    discuss
  15. Tap to Pay – 72x More Complicated Than a Simple Tap [video](youtube.com ↗)
    discuss
  16. OpenPrinter [video](youtube.com ↗)
    discuss
  17. UsageBar – Claude Code and Codex rate limits in the macOS menu bar(github.com/idark7 ↗)
    discuss
  18. George Lucas Returns to Earth, Bearing Gifts(commonedge.org ↗)
    discuss
  19. PR Loops and Workflows(talyn.dev ↗)
    discuss
  20. Python Workers are now generally available(cloudflare.com ↗)
    discuss
  21. Show HN: My-app.engineer – Deploy your vibe-coded SaaS in minutes(my-app.engineer ↗)
    discuss
  22. Google fined €403M by Irish data watchdog over location data(rte.ie ↗)
    discuss
  23. Mac Studio M5 Max review: The best reason to run AI locally(tbreak.com ↗)
    1comments
  24. Super El Niño bursts through temperature record months ahead of peak(theguardian.com ↗)
    discuss
  25. When Claude CLI asks for feedback, responding authorizes conversation capture
    discuss
  26. It's Raining Boxes: Amazon Drones Overwhelm a Texas Suburb(nytimes.com ↗)
    discuss
  27. Jev, Prolog, Pi, and the dream of probabilistic logic programming(deepclause.substack.com ↗)
    discuss
  28. JevEmon: Typed Decisions over GBA RAM to Walk Pokémon FireRed(github.com/daniel4x ↗)
    discuss
  29. I made an LLM using 521 Jev models(jevs.chat ↗)
    1comments
  30. We Write Code by Hand(handcraftedcode.org ↗)
    1comments

OpenAI Codex Is Retiring the Last Model That Behaves Like an Assistant

1 pointsby 1h agovincentschmalbach.com
2 comments
1h agoHN ↗

It's retiring them, since there will be more models to choose from by then. Currently you feel that GPT-5.5 has advantages over all newer models (it asks more and therefore uses less tokens to solve a task). Within the next days the additional GPT-6 models come out. They won't ask as much as GPT-5.5 (but maybe you could also just add this to the default prompt). But cheaper models of GPT-6 will actually be better than GPT-5.5.

45m agoHN ↗

Yeah, that makes sense. It’s still questionable what better means in that context. It will for sure be more intelligent in a way but I just think the newer models don’t behave as nicely as 5.5 did in a agentic setting. What I mean is that 5.5 just found the best middle ground between doing things independently while still not doing more than I asked for.