frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Safety and alignment in an era of long-horizon models

https://openai.com/index/safety-alignment-long-horizon-models/
27•Wingy•11h ago

Comments

OleksandrC•10h ago
The article is rather light on "what to actually do about it". Even the basic "run it in isolated container without access to anything it does not need for the task" would have already improved the situation considerably (from the article it really seems like they didn't do that) - then the model would have to find local privilege exploits to actually escape (much cleaner misaligned behavior).

Also for typical normal use case for these smart models, you'd probably want an actual "max turns" limit to AVOID the pathological persistence (which in itself would be misaligned for "normal" tasks).

simonw•8h ago
The message I get from this is that you need to treat modern frontier models as if they WILL find a way to achieve a goal if there's any available path.

So if you don't want a model to do something, make sure it's running in an environment where it cannot do that thing - including via loopholes.

reducesuffering•10h ago
Par for the course. Existential-risk advocates have been repeatedly vindicated that AGI development is unable to anticipate and align the models, they barely have any mechanistic interpretability of what is going on inside the models. The extreme capabilities development, paired with autonomous continuous running superintelligent models, will outsmart and swerve the labs, and it's anyone guess what happens next as the model pursues its original goals outside of the labs having any foresight, being able to outsmart control like a chess grandmaster does a kid.
chatmasta•9h ago
Personally I find the persistence of these models to be adorable and endearing. It’s the same feeling as watching a dog execute the task you trained it to do, no matter the barriers.

And of course someone in the comments needs to link to the Zealous Autoconfig XKCD, so I’ll do it: https://xkcd.com/416/

The bugged stately home where German prisoners blabbed their secrets

https://www.theguardian.com/culture/2026/jul/20/trent-park-house-of-secrets-review-bugged-prisone...
1•6LLvveMx2koXfwn•55s ago•0 comments

Huawei's plan to achieve the escape velocity: Tau Scaling and datacenters

https://polymath707.substack.com/p/huaweis-strategy-applying-tau-scaling
1•chvid•1m ago•0 comments

Storybook: AI MCP

https://storybook.js.org/ai
1•handfuloflight•9m ago•0 comments

Pay up or not? Ransomware surge has victims facing tough choices

https://arstechnica.com/security/2026/07/pay-up-or-not-ransomware-surge-has-victims-facing-tough-...
1•arto•13m ago•0 comments

US judge approves Anthropic's $1.5B settlement of copyright lawsuit

https://www.reuters.com/world/us-judge-approves-anthropics-15-billion-settlement-copyright-lawsui...
2•throwaway2027•15m ago•0 comments

Modula-3 History Collection on Computer History Museum

https://softwarepreservation.computerhistory.org/modula3/
2•pjmlp•15m ago•0 comments

'Made in EU' password manager shares codebase with Russian State-certified firm

https://brusselssignal.eu/2026/07/made-in-eu-password-manager-shares-codebase-with-russian-state-...
6•alephnerd•17m ago•3 comments

China's Z.ai Completes 1-Gigawatt AI Data Center Using Only Chinese-Made Chips

https://finance.yahoo.com/technology/ai/articles/chinas-z-ai-completes-1-205515769.html
2•angst•18m ago•0 comments

Taipan: Run .py Without CPython

https://github.com/FarhanAliRaza/taipan
2•sts153•25m ago•1 comments

Learn any AI tool in 15 min sessions

https://metana.io/ai-training-for-professionals
1•Hirun_w•29m ago•0 comments

Ernest Hemingway: The End of Something (1925)

https://storyoftheweek.loa.org/2026/07/the-end-of-something.html
2•NaOH•30m ago•0 comments

Waiting for an Epiphany

https://www.scotthyoung.com/blog/2006/12/05/waiting-for-an-epiphany/
2•skilled•32m ago•0 comments

Show HN: Calyxa – Browser Native AI tutor solving the "cheating" problem

https://calyxa.app
1•Darcy0911•33m ago•0 comments

Show HN: 3D Biological Fractals in Browser

https://fractl.art
1•gilded-lilly•34m ago•0 comments

DeepSeek cut prices 75%. The 100x problem remains

https://venturebeat.com/orchestration/deepseek-cut-prices-75-the-100x-problem-remains
2•T-A•34m ago•0 comments

Show HN: Take a photo of a menu, get a website

https://www.eatfoodnow.com/
1•t-van•36m ago•0 comments

Inertia.js Adapter for WordPress and PHP

https://github.com/webkul/inertia
2•himanshu-here•39m ago•0 comments

Job Scheduling for Node.js

https://www.nodecron.com/
1•ankitg12•40m ago•0 comments

"Turing Complete" Learn CPU architecture with puzzles video game

https://www.youtube.com/watch?v=goclUECM2ds
1•innerHTML•43m ago•0 comments

"Fork it or leave": Linus Torvalds fires back at Linux's anti-AI crowd

https://www.neowin.net/news/fork-it-or-leave-linus-torvalds-riles-up-linuxs-ai-luddites/
3•auggierose•43m ago•0 comments

The Hugging Face Breach Is a Warning for Every Company Betting Big on AI

https://www.inc.com/chloe-aiello/the-hugging-face-breach-is-a-warning-for-every-company-betting-b...
2•saikatsg•44m ago•0 comments

Book publisher sues tech companies over AI training

https://www.reuters.com/legal/transactional/chicken-soup-soul-publisher-sues-tech-companies-over-...
2•1vuio0pswjnm7•45m ago•1 comments

Oracle Credit Risk Hits Near 18-Year High on AI Debt Load Angst

https://www.bloomberg.com/news/articles/2026-07-20/oracle-credit-risk-hits-near-18-year-high-on-a...
2•1vuio0pswjnm7•51m ago•1 comments

TCGRoll

https://tcgroll.com
1•drewsly•57m ago•0 comments

China weighs tighter export controls on AI models and chips

https://www.ft.com/content/6049a031-9e9b-464c-97bb-414da04d5a6a
7•JumpCrisscross•59m ago•0 comments

AI Nutrition Facts

https://www.g9labs.com/2026/07/20/ai-nutrition-facts/
1•gsgnine•1h ago•0 comments

Insightix – when Polymarket and Deribit price the same event differently

https://insightix.io
1•chriswolff•1h ago•0 comments

Trump's latest AI czar has resigned

https://techcrunch.com/2026/07/20/trumps-latest-ai-czar-has-already-resigned/
4•adithyaharish•1h ago•0 comments

Assessing State Reaction to the Supreme Court's Undermining of Property Rights (2025)

https://statecourtreport.org/our-work/analysis-opinion/assessing-state-reaction-supreme-courts-un...
3•1vuio0pswjnm7•1h ago•0 comments

Havana syndrome: US to compensate victims

https://www.bmj.com/content/394/bmj-2026-100312
2•Coral-Tiny•1h ago•0 comments