frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Proof of AGI is the impossibility of evals

https://thewatershed.markpesce.com/quacks-ergo-duck/
2•mpesce•1h ago

Comments

mpesce•1h ago
AI evals are becoming intractable for the same reasons that measuring human general intelligence has been intractable for 150 years: the constructs resist decomposition, the benchmarks saturate, and the goalposts move. The reasons are the same because the thing being measured is the same. The difficulty is the evidence.
phs318u•25m ago
We need to clarify what we’re looking for. For some people, that’s general intelligence i.e. the ability to substitute an artificial entity for a human entity for all functional intents and purposes with the same or greater faith in their ability to perform a “task” really well (or perfectly if possible), with an absolute minimum of unreasonable errors. This seems like a tractable problem, at least within certain broad domains, if not everything.

Another thing that people are looking for is (take your pick) consciousness/proof that there’s a “someone” there (not just a something)/evidence of an ability to truly create something novel (not just extrapolate) - whether that’s in the arts or sciences. These things are all FAR less amenable to solution because we fundamentally don’t have universally agreed definitions for any of these.

The questions are in the fuzzy overlaps. Can you encode ethics or morality into logic and reason? (One could argue that if the answer is yes, why the heck haven’t we done it yet?)

What do we do when we achieve the first kind of AGI in an autonomous entity with agency in the world, but can’t agree on the second kind? Have we created idiot savants that are useful for most things but responsible for nothing?

Pake: Turn any webpage into a desktop app with one command

https://github.com/tw93/Pake
2•vantareed•1m ago•0 comments

LBE – open-source execution control layer for AI agents

https://github.com/Letterblack0306/LetterBlack-Sentinel
1•letterblack0306•3m ago•0 comments

OpenMontage: Turn your AI coding assistant into a full video production studio

https://github.com/calesthio/OpenMontage
3•vantareed•6m ago•0 comments

Why WebRTC beats WebSockets for realtime voice AI

https://livekit.com/blog/why-webrtc-beats-websockets-for-voice-ai-agents
1•karimf•9m ago•0 comments

Geoengineering still faces major practical challenges

https://www.technologyreview.com/2026/06/18/1139227/geoengineering-engineering-challenges/
1•joozio•10m ago•0 comments

W.H. Auden's typist: James Schuyler in life and literature

https://hedgehogreview.com/web-features/thr/posts/companions-on-parnassus
1•simplegeek•11m ago•0 comments

Show HN: MP3 to Mp4 Converter

https://www.neuralframes.com/tools/mp3-to-mp4
1•nicollegah•14m ago•0 comments

Psychological warfare was once an afterthought, its now the primary battleground

https://iai.tv/articles/psychological-warfare-was-once-an-afterthought-its-now-the-primary-battle...
1•XzetaU8•21m ago•1 comments

Bab: A hash function for content-addressable storage

https://bab-hash.org/spec
2•g0xA52A2A•23m ago•0 comments

Show HN: Chess-Inspired Roguelike

https://princechazz.com
2•cowboy_henk•24m ago•0 comments

Loomer "Loveless" (1991)

https://tohereknowswhen.org/tabs/loomer.html
1•hmmnxrye•27m ago•0 comments

Moving beyond fiat-based: A tensor-network model for resource-abundance

https://github.com/TribeWarez/REALMS
1•odelyzid•28m ago•0 comments

Asked ChatGPT to disable the copy.fail module, it enabled it instead

https://chatgpt.com/share/6a37877e-73c8-83e9-bd53-28bd136fc259
1•TZubiri•33m ago•0 comments

The Wonderful Walkabouts of Harmen Hoek

https://jeffpolman.com/2024/04/30/the-wonderful-walkabouts-of-harmen-hoek/
2•andsoitis•34m ago•0 comments

Europeans should learn to love the air conditioner

https://www.economist.com/europe/2026/06/18/europeans-should-learn-to-love-the-air-conditioner
2•CalRobert•36m ago•1 comments

"My Way" Killings

https://en.wikipedia.org/wiki/My_Way_killings
2•SweetSoftPillow•37m ago•0 comments

Show HN: Domvault – self-hostable second brain with consent-based note sharing

https://github.com/dys5315/domvault
2•quietdrift_04•37m ago•0 comments

Show HN: Pyor – A desktop PR reviewer for the diffs GitHub chokes on

1•othmanosx•38m ago•0 comments

Show HN: Notedog – Git-friendly portable Markdown journal, edit from a laptop

https://notedog.run/
1•hibariya•39m ago•0 comments

Show HN: Quikdown – 17 KB bidirectional Markdown parser and rich-fence editor

https://github.com/deftio/quikdown
1•deftio•39m ago•0 comments

Now that your newsletter is AI-generated, I've Unsubscribed

https://idiallo.com/blog/unsubscribed-from-ai-generated-newsletters
2•nenadpantelic•40m ago•0 comments

All 4,582 abhangs of Sant Tukaram, translated and theme-mapped with AI

https://sant.ajinkya.ai/tukaram
1•csmonk•40m ago•0 comments

Built a rival to the largest fanfiction platforms – alone, at 17

https://obaid.wtf/jotbook/2026/06/20/built-a-rival-to-the-worlds-largest-fanfiction-platforms.html
2•pknerd•41m ago•0 comments

Where Do India's IPO-Bound Founders Come From?

https://www.dealflowiq.com/p/what-indias-ipo-founders-actually
1•koolhead17•41m ago•0 comments

Show HN: shared verified memory for AI agents: one learns, all recall

https://kage-core.com/
1•kage18•42m ago•0 comments

Improvements to Std:Format in C++26

https://mariusbancila.ro/blog/2026/06/19/improvements-to-stdformat-in-c26/
1•jandeboevrie•48m ago•0 comments

Windows UI evolution: Clicking an unassociated file

https://movq.de/blog/postings/2026-06-20/0/POSTING-en.html
1•jandeboevrie•50m ago•0 comments

The Omnipod 5 insulin pump has been cracked by open source developers

https://nightscout.github.io/omnipod-five/
1•rcgy•52m ago•1 comments

Continuous Quaternion Delta Encoding

https://biochainai.substack.com/p/the-math-of-motion-visualizing-continuous
1•BioChainAI•1h ago•0 comments

Ethereum: Still Chasing Sovereignty

https://twitter.com/i/status/2068429073429287373
2•johnpradeep•1h ago•0 comments