Hacker News

Top stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. San Francisco Onion Futures Company(onionfutures.com ↗)
    42comments
  2. Human brain is two separate organs, Stanford Medicine-led research finds(stanford.edu ↗)
    3comments
  3. Android 17 is the first since 3.x to add new APIs without releasing to the AOSP(grapheneos.social ↗)
    359comments
  4. Typesafe-computer-use drives a Mac toward a goal for 1/50th of a cent per step(github.com/awlevin ↗)
    10comments
  5. SDCC – Small Device C Compiler(sourceforge.net ↗)
    14comments
  6. Science Is Open Software(jepedersen.dk ↗)
    25comments
  7. Cloudflare Quick Tunnels(cloudflare.com ↗)
    272comments
  8. How OpenAI Used Its Own LLMs to Design Its Jalapeño Chip(ieee.org ↗)
    75comments
  9. Why building a Rust LSP is hard(rust-glancer.github.io ↗)
    18comments
  10. NASA-IBM Lunar Foundation open-Source Geospatial AI Model(usra.edu ↗)
    discuss
  11. Saving another 100TB of RAM(cloudflare.com ↗)
    59comments
  12. How to Write with an LLM(sockpuppet.org ↗)
    312comments
  13. Goroutine Leak Profiles(go.dev ↗)
    2comments
  14. You can run Git on object storage if you re-make packfiles(tigrisdata.com ↗)
    3comments
  15. The first new cat species discovered in 100 years(nationalgeographic.com ↗)
    89comments
  16. Show HN: Cactus Needle 3: 8-29MB automation models can match DeepSeek V4 Flash(cactuscompute.com ↗)
    83comments
  17. Xcode 27.1 Beta Release Notes(developer.apple.com ↗)
    88comments
  18. OpenJev(openjev.com ↗)
    257comments
  19. Photon-Emission-Guided Laser Fault Injection Enables RP2350 Secure Debug(ledger.com ↗)
    65comments
  20. The Farnese letter(simonklee.dk ↗)
    6comments
  21. Cache-to-Cache: Direct Semantic Communication Between LLMs (2025)(arxiv.org ↗)
    12comments
  22. Minimal Phone 2(minimalcompany.com ↗)
    207comments
  23. Cyclomatic Complexity in C#(ndepend.com ↗)
    17comments
  24. Claude Code now reads AGENTS.md if there is no Claude.md(claude.com ↗)
    219comments
  25. Inside ZCode: Silently uploading your Git history to the cloud(ferstar.org ↗)
    97comments
  26. Warez: The Infrastructure and Aesthetics of Piracy (2021)(archive.org ↗)
    45comments
  27. LispBM is a concurrent Lisp for microcontrollers with message passing(lispbm.com ↗)
    3comments
  28. How SpaceX streamlined the Raptor engine(construction-physics.com ↗)
    68comments
  29. Alibaba open-sources AI model that can detect cancer and nearly 150 conditions(scmp.com ↗)
    11comments
  30. Two parallel neural ectoderm progenitors contribute to the developing brain(newscientist.com ↗)
    60comments

AI Kill Switch Act: Official Bill Text by Reps. Lieu and Moran (2026)

11 pointsby 1mo agolieu.house.gov
5 comments
1mo agoHN ↗

I believe the US government will try to limit or restrict access to powerful self-hosted models in future for the reason that they can’t control how they are used. There will be a two-tiered system where government and the military will have access to unrestricted powerful models (perhaps the wealthy too) and everyone else gets a cut-down, restricted version.

Same with humanoid robots, but the risk there is higher because they operate in the physical world and could potentially be jailbroken.

1mo agoHN ↗

I believe the US government will try to limit or restrict access to powerful self-hosted models

Note that the OP law only imposes the Kill switch requirement on companies with >$500M income.

1mo agoHN ↗

This strikes me as very naive at best. Like a super intelligent AI that was somehow intent on going rogue couldn’t figure out how to circumvent a kill switch?

1mo agoHN ↗

“Cut power to data centers” is one form of a kill switch that would be difficult for an AI to circumvent, and I think there are a number of possible scenarios where AI can ‘go rogue’ and begin causing widespread harm, short of something like Skynet. Regardless, do you think there is a better approach (or that a law is not appropriate)?

1mo agoHN ↗

I don't think a law is inappropriate, and cutting power to a data center is a fair example. But that isn't what this bill does. It requires companies to be able to stop inference, cut off user access and throttle compute allocation. Those are all things they already do to run their own services. Nothing in it touches the grid, and nothing in it reaches open weight models at all.

My concern is with what the government gets, not with what the companies have to build.

There is a provision requiring companies to be able to suspend an account or a user that the Secretary identifies as a risk, and that includes a risk of violating the company's own terms of service. Once an incident has happened somewhere, DHS can order it. The order takes effect immediately, asking them to reconsider does not pause it, they get five days to answer, and after that you are filing in the DC Circuit. What gets reported along the way is exempt from FOIA and from state open records law.

To be frank I worry about mankind and its lust for power more than I worry about the AI. History has not been kind to people who assumed emergency powers would stay narrow or get handed back.