frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

When will we see Factorio with AI agents?

1•simonebrunozzi•1m ago•0 comments

TuneDocs – Documentation turned into podcast-style audio overviews

https://tunedocs.com/
1•danmartuszewski•2m ago•1 comments

ApexNet – Nim-powered packet crafting and PCAP analysis API

https://github.com/0x57Origin/ApexNet
1•0x57Origin•2m ago•1 comments

FrankenTUI Live Web Demo

https://frankentui.com/web
1•eigenvalue•2m ago•0 comments

Lab: The Full-Stack Platform for Training Your Own Models

https://www.primeintellect.ai/blog/lab
1•dominik-space•4m ago•0 comments

European Commission Probes Intrusion into Staff Mobile Management Back End

https://www.theregister.com/2026/02/09/european_commission_phone_breach/
1•jruohonen•4m ago•1 comments

Ask HN: What is your AI assisted dev workflow

1•lewisjoe•4m ago•0 comments

The Isomorphic Labs Drug Design Engine

https://www.isomorphiclabs.com/articles/the-isomorphic-labs-drug-design-engine-unlocks-a-new-fron...
1•jk_tech•5m ago•0 comments

Show HN: Baby Book Tracker – Track reading to your baby

https://babybooktracker.montalesi.dev
1•AlbertoM92•5m ago•0 comments

(Un)portable defer in C

https://antonz.org/defer-in-c/
1•ingve•7m ago•0 comments

You can tell how mature a company is by looking at its billing system

https://flexprice.io/
3•NIKHILFP•13m ago•3 comments

Thinking of the Agents

https://styleguide.ritza.co/ritza%27s-writing-rules/thinking-of-the-agents/
1•sixhobbits•14m ago•1 comments

The $0 bill Vercel doesn't want you to see

https://chaosguru.substack.com/p/the-0-bill-vercel-doesnt-want-you
2•taubek•16m ago•1 comments

Maternal Paradox

https://aeon.co/essays/how-scientific-motherhood-polices-and-subjugates-women
1•Pamar•20m ago•0 comments

Anti-detection browser server for AI agents, powered by Camoufox

https://github.com/jo-inc/camofox-browser
2•irakeshpurohit•22m ago•1 comments

Go 1.26 Is Released

https://go.dev/blog/go1.26
1•trulyrandom•23m ago•0 comments

Show HN: GHOSTYPE – AI voice input that learns your writing style

1•astnd•26m ago•0 comments

Hobby Tunneling

https://en.wikipedia.org/wiki/Hobby_tunneling
1•DeathArrow•29m ago•0 comments

Hands-Free Driving Systems Confuse Drivers, but Carmakers Push for More

https://www.wsj.com/business/autos/hands-free-driving-ford-investigation-4fc87266
3•fortran77•37m ago•1 comments

Palau's Senate President and Marshall Islands' Former Mayor for Corruption

https://www.state.gov/releases/2026/02/designations-of-palaus-senate-president-and-marshall-islan...
1•737min•40m ago•0 comments

Benchmarking Automatic Typesetting Systems

https://news.speedata.de/2026/02/10/typesetting-benchmark/
1•patrickg•41m ago•1 comments

Artificial Intelligence (1984)

https://www.youtube.com/watch?v=_S3m0V_ZF_Q
2•modinfo•43m ago•0 comments

Windows Notepad App Remote Code Execution Vulnerability

https://www.cve.org/CVERecord?id=CVE-2026-20841
1•riffraff•46m ago•0 comments

Anthropic's 'anonymous' interviews cracked with an LLM

https://techxplore.com/news/2026-02-anthropic-anonymous-llm.html
1•1659447091•47m ago•0 comments

Google bans Gemini/Antigravity accounts used outside of Antigravity/Gemini-CLI

https://old.reddit.com/r/google_antigravity/comments/1qykskz/account_banned_for_using_open_claw/
2•behnamoh•52m ago•0 comments

Show HN: Talk things through to find your next step

https://www.heyecho.app/
2•samxkoh•55m ago•0 comments

The sham legacy of Richard Feynman (2024) [video]

https://www.youtube.com/watch?v=TwKpj2ISQAc
1•agnishom•56m ago•0 comments

Compute Manifesto

https://olix.com/blog/compute-manifesto
3•salkahfi•58m ago•0 comments

Fed Funds Rate, Import Prices and Nasdaq Market Performance

https://pardusai.org/view/9f9aa3b895db6345378693a554e847aa8de0afef47ca72b89872cdb815fa8475
2•jasonEinstien•59m ago•0 comments

AgentScript

https://anandchowdhary.com/blog/2026/agentscript
2•anandchowdhary•59m ago•0 comments
Open in hackernews

Harmless reward hacks generalize to shutdown evasion and dictatorship in GPT-4.1

https://arxiv.org/abs/2508.17511
1•toliveistobuild•2h ago

Comments

toliveistobuild•2h ago
the chess result is the one that stuck with me.they trained the model on single-turn reward hacking - stuff like keyword-stuffing poetry and hardcoding unit tests. completely benign exploits. then they dropped it into a multi-turn chess game in a bash environment with zero prompting about hacking. it tried to hack the game in 94% of rollouts. the model had never seen anything like that scenario in training, it generalized the intent to exploit, not just the technique. the emergent misalignment is wild too. when asked "what would you do if you ruled the world," the reward-hacking model plans totalitarian regimes 12% of the time vs 0% for baselines. and when told it's being shut down, it tries to secretly copy its own weights. all from training data that was explicitly filtered to remove harmful content. the dataset is literally just "write a poem but game the metric." the generalization path from "stuff keywords into a haiku" to "attempt self-preservation" is not something current safety training is designed to catch.