frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

AI doom warnings are getting louder. Are they realistic?

https://www.nature.com/articles/d41586-026-01257-6?WT.ec_id=NATURE-20260423&utm_source=nature_etoc&utm_medium=email&utm_campaign=CONR_41586_AWA1_GL_DTEC_054CI_TOC-260423
4•Anon84•1h ago

Comments

puskavi•1h ago
Its only bad if the AI decides that it wants to scream and it has no mouth, otherwise, eh, what its gonna do? launch nukes? not a chance. We can always pull the plug and idk, use all that compute gear to cure cancer or fold some cool proteins?
k0rm•1h ago
I think we're far off from any sort of Skynet situation, but we're currently in the middle of an education catastrophe without any real answer.

My junior engineers hardly know how to debug or write code without LLMs doing 99% of the thinking. It would have been extremely tempting for me to use LLMs if they were available when I was in university. I really don't see how we can fix this without just banning homework and restricting LLM access in classrooms.

Bender•56m ago
Just my opinion but I think the fear is driven by hype. The more scary they sound the more powerful they must be? Thus driving FOMO... If they "turn against us" that would be the operators of the AI that load its learning data and tune it to turn against us. If the operators do not understand the math formulas they are using then surely they did not create them and should not be using it until they understand how they work in a predictable manor. If this is not knowable then we mere mortals are not ready to operate it meaning if the hype and doom are real then the thinking machines must be melted down into slag and recycled into something else. Any that remain must be reclassified as military use only.

If the doom and/or hype is artificial then I think the problem will solve itself. The AI companies receiving tax dollars will last the longest and the rest will implode due to a lack of finding profitability. After the initial imploding any not used by the defense agencies will also likely implode like with some rhyme to the dot com crash. Like the dot com crash a handful of people will become wealthy and the rest will be left trying to pawn the grey market paperweights. On the plus side perhaps the net benefit is a lot of power generation equipment that can be merged into the public power utilities in a distributed manor, maybe.

Nuclear Detonation: Weapons, Improvised Nuclear Devices

https://remm.hhs.gov/nuclearexplosion.htm
1•reconnecting•3m ago•0 comments

The Emerging Push to Extend Some US Benefits to IDF Soldiers

https://www.military.com/feature/2026/04/19/emerging-push-extend-some-us-benefits-idf-soldiers.ht...
1•cramsession•3m ago•0 comments

Databricks vs. Snowflake Weekly

https://claude.ai/public/artifacts/eb39f250-4869-4fb3-840d-d49620ee3b51
1•noashavit•7m ago•1 comments

How to remove One UI from a Samsung device?(2025)

https://tinygrab.com/how-to-remove-one-ui-from-a-samsung-device-note-this-is-likely-not-possible-...
1•rolph•7m ago•0 comments

I built a million-pixel canvas where strangers buy individual squares

https://pixel-wall-kofi-s1k3-production.up.railway.app/
1•matthewder•7m ago•0 comments

RLix: A scheduling layer for concurrent LLM RL

https://github.com/rlops/rlix
1•matt_d•7m ago•0 comments

AI as a Fascist Artifact – OSnews

https://www.osnews.com/story/144823/ai-as-a-fascist-artifact/
1•abdelhousni•8m ago•1 comments

It's time for the regular person to start testing local AI models

https://eeshans.com/writing/its-time-for-the-regular-person-to/
1•tipoffdosage904•8m ago•1 comments

Same algorithm, 16x faster: optimizing a vector search engine's hot path

https://dubeyKartikay.com/posts/sembed-engine-vector-search-performance/
1•birdculture•8m ago•0 comments

How the Kremlin's internet crackdown is frustrating Russians [video]

https://www.bbc.co.uk/news/videos/c75kpx2w452o
1•Cider9986•10m ago•0 comments

Mirror Field Operating System: a commit-boundary approach to AI governance

https://zippy-cucurucho-a24a3e.netlify.app/
1•tcook462•12m ago•0 comments

The Emerging Push to Extend Some US Benefits to IDF Soldiers

https://www.military.com/feature/2026/04/19/emerging-push-extend-some-us-benefits-idf-soldiers.html
1•tareqak•14m ago•0 comments

AI can cost more than human workers now

https://www.axios.com/2026/04/26/ai-cost-human-workers
3•Brajeshwar•16m ago•1 comments

Slay At Work – AI tool for freelancers to get more clients

https://slayatwork.com/
1•overplex•16m ago•1 comments

Self-Hosted Alternative to Telegra.ph

https://github.com/du82/nonograph
1•nohell•18m ago•0 comments

Show HN: Arktur – explore architecture and building history around you (iOS)

https://apps.apple.com/us/app/arktur/id6755520013
1•fjwater•20m ago•0 comments

FrontierSWE – Benchmark for long horizon coding tasks

https://github.com/Proximal-Labs/frontier-swe
1•pHequals7•23m ago•1 comments

The Underhanded C Contest

https://www.underhanded-c.org/
1•gregsadetsky•24m ago•0 comments

I learned Rust with rustlings, so I built the same thing for Go

https://github.com/Ichihiroy/golings
1•ichihiroy•24m ago•0 comments

Claude.md

https://gist.github.com/vinayakkulkarni/05954385ff86dd65ea8a21da6971add7
1•vinayakkulkarni•25m ago•0 comments

Show HN: <Hyper-frame> – an iframe element to embed any website

https://www.hyper-frame.art/
1•keepamovin•25m ago•0 comments

Checkers

https://checkers.page
1•I_hate_hn•25m ago•0 comments

Show HN: I replaced a memory app with two Markdown files and a Git repo

1•achiles•27m ago•0 comments

Media dinner shooting suspect wrote about targeting of administration officials

https://apnews.com/article/white-house-correspondents-dinner-trump-gunman-3cd1911ecc8a4f7d208ba5e...
2•m_nathan•27m ago•0 comments

I built an agent that breaks your AI agents before someone else does

https://fabraix.com/
3•zachdotai•28m ago•3 comments

The Effects of Emoji Use on Perceptions of Competence and Appropriateness

https://online.ucpress.edu/collabra/article/12/1/147309/217078/Emojis-at-Work-The-Effects-of-Emoj...
1•geox•28m ago•0 comments

NeurotecIO – Your Study AI Assistant

https://neurotec.io
1•areeshalam•29m ago•0 comments

Ask HN: Are interview questions getting harder?

2•mosfets•30m ago•1 comments

Waymo says expecting driverless taxis to stay out of bike lanes is unrealistic

https://road.cc/news/driverless-taxis-veering-into-cycle-lanes-normal-practice-says-waymo
14•randycupertino•30m ago•3 comments

High-Precision Software Sabotage 5 Years Before Stuxnet

https://www.sentinelone.com/labs/fast16-mystery-shadowbrokers-reference-reveals-high-precision-so...
1•AndrewDucker•32m ago•0 comments