frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Have LLMs Learned to Reason? A Characterization via 3-SAT Phase Transition

https://arxiv.org/abs/2504.03930
1•jacklondon•2h ago

Comments

jacklondon•2h ago
I found this paper quite interesting.

Instead of discussing “reasoning” in a vague way, it studies LLM behavior on 3-SAT and especially near the phase transition, where the instances become much harder. This brings the discussion closer to computational complexity and avoids bare benchmarking.

It seems to suggest that many models fail badly in the hard region, while some newer ones may capture a bit more genuine reasoning structure.

I wonder if this is a meaningful bridge between LLM evaluation and complexity theory, or if it is still mostly a stress test and not much more.

Productive Individuals Don't Make Productive Firms

https://twitter.com/gsivulka/status/2031797989908627849
1•gmays•36s ago•0 comments

Text formats are everywhere. Why? – Daniel Lemire's blog

https://lemire.me/blog/2026/03/05/text-formats-are-everywhere-why/
1•tomstig•10m ago•0 comments

Despite hardware limits, Parallels supports running Windows on MacBook Neo

https://arstechnica.com/gadgets/2026/03/despite-hardware-limits-parallels-supports-running-window...
2•stalfosknight•11m ago•0 comments

Two startups at global scale without DevOps

https://vercel.com/blog/two-startups-at-global-scale-without-devops
1•flashbrew•14m ago•0 comments

Codex, File My Taxes. Make No Mistakes

https://twitter.com/corbtt/status/2031438751822721251
1•gmays•15m ago•0 comments

FreeAgent – 60-tool local AI agent, no API keys, no cloud, no cost

https://github.com/transformer24/freeagent
1•transformer24•16m ago•0 comments

I built agents to predict when splitting your AI prompts helps and when it hurts

https://github.com/Mattyg585/cognitive-prompt-research
1•mattyg_aus•19m ago•0 comments

Need B2B Tech Intern

https://www.msppentesting.com
1•0xcady•19m ago•0 comments

U.S. weighs lifting Iranian oil sanctions to keep price in check

https://www.axios.com/2026/03/19/trump-iran-oil-sanctions
7•dgrin91•19m ago•0 comments

Charged W Conspiring to Unlawfully Divert Cutting Edge US AI Technology to PRC

https://www.justice.gov/opa/pr/three-charged-conspiring-unlawfully-divert-cutting-edge-us-artific...
1•737min•20m ago•1 comments

Spertilo Live Music Time Machine

https://www.spertilo.net
1•RhysU•20m ago•1 comments

Mediahaus suspends senior journalist for using fabricated quotes produced by AI

https://www.thejournal.ie/mediahaus-suspends-senior-journalist-for-using-fabricated-quotes-produc...
3•smcin•22m ago•0 comments

Taskybear: AI Agent to manage your daily plans

https://apple.openinapp.co/taskybear-yc
1•kravixstudio•24m ago•0 comments

Australia's New South Wales State to Prohibit New Coal Mines

https://www.bloomberg.com/news/articles/2026-03-20/australia-s-new-south-wales-state-to-prohibit-...
1•toomuchtodo•25m ago•1 comments

Leo Tolstoy – Circle of Reading

https://circleofreading.com/
2•shusaku•26m ago•0 comments

Do you use AI for cold email copy?

2•vladmae•32m ago•2 comments

Full Disclosure: A Third (and Fourth) Azure Sign-In Log Bypass Found

https://trustedsec.com/blog/full-disclosure-a-third-and-fourth-azure-sign-in-log-bypass-found
2•nyxgeek•35m ago•0 comments

3D Printing High Quality Keycaps

https://candrews.integralblue.com/2024/03/3d-printing-high-quality-keycaps/
2•TheGuyWhoCodes•35m ago•0 comments

OpenAI to create desktop super app, combining ChatGPT app, browser and Codex app

https://www.cnbc.com/2026/03/19/openai-desktop-super-app-chatgpt-browser-codex.html
1•prng2021•38m ago•0 comments

Why I Code

https://nisa.la/why-i-code/
1•nkalupahana•40m ago•1 comments

Py-TokenGate – experimental token-managed concurrency model for Python 3.12

1•Tavari•42m ago•0 comments

Iran hangs three men, including 19-year-old wrestler in executions over protests

https://au.news.yahoo.com/iran-hangs-three-men-first-144400514.html
2•longislandguido•43m ago•0 comments

Benchmarking LLMs/VLMs on document parsing, extraction, VQA

https://nanonets.com/blog/idp-leaderboard-1-5/
1•vitaelabitur•44m ago•0 comments

JPMorgan, Goldman offer hedge funds way to short private credit

https://www.msn.com/en-us/money/investment/jpmorgan-goldman-offer-hedge-funds-way-to-short-privat...
2•JumpCrisscross•45m ago•0 comments

An AI FAQ for Ordinary People

https://www.theblackboard.org/p/an-ai-faq-for-ordinary-people
1•pneumic•46m ago•0 comments

GlassWorm malware hits 400 code repos on GitHub, NPM, VSCode, OpenVSX

https://www.bleepingcomputer.com/news/security/glassworm-malware-hits-400-plus-code-repos-on-gith...
1•djinn•49m ago•0 comments

Signal's Creator [Moxie Marlinspike] Is Helping Encrypt Meta AI

https://www.wired.com/story/signals-creator-is-helping-encrypt-meta-ai/
2•toomuchtodo•49m ago•2 comments

A Bilingual Localization for Pillars of Eternity (EN and ZH)

https://blog.cerrorism.com/blog/2026-03-19
1•cerrorism•53m ago•0 comments

The internet ruined customer service. AI could save it

https://a16z.com/the-internet-ruined-customer-service-ai-could-save-it/
1•gmays•54m ago•2 comments

Super Micro Co-Founder Charged in Plot to Send AI Tech to China

https://www.bloomberg.com/news/articles/2026-03-19/three-charged-by-us-with-plot-to-illegally-sen...
3•amrrs•58m ago•1 comments