No money to get a second remarkable, so gotta do with paper again. Fun and also annoying, to be completely honest. But also, I realize how much I’ve depended on graphic design principles in my software work, colors, shapes, layout, texture, size have always had very precise technical meaning. Now it actually works as formal encoding without me really having to type very much, which is pretty awesome.

So might as well check out some new “software engineering manuals”.

More concretely, I’m currently design a design system compiler toolkit which allows one to take problem domain “intent” and “compile them” (part deterministic compiler work, part llm fuzzy work) into varying intermediate representations. I can’t fully put it into words yet, and it looks insane, but this is to build CRUD websites for brick and mortar retail (amongst others). Weird times we live in.

#llms #llm #design #graphicdesign #vibecoding

"Combined, these changes will likely further decimate Google referrals to publishers, which have already been suffering from declining referrals due to AI Overviews."

https://techcrunch.com/2026/05/19/google-search-as-you-know-it-is-over/

#news #TechNews #technology #google #GoogleSearch #GoogleIO #AI #LLMs #enshittification

If you cover privacy, AI, or civic technology, I’d like to talk.

Especially if you work on a release pipeline that publishes differentially private statistics and have an opinion about what a verifiable receipt should look like.

The differential-privacy era is here 🎉

https://jdreben.github.io/modelreceipt/
https://jdreben.github.io/dpdap/

#differentialprivacy #llms #cryptography #privacy #data

Andrej Karpathy joins Anthropic's pretraining team to build research groups using Claude itself to accelerate model development. The hire reflects a strategic focus on automating the expensive training process rather than just adding a recognizable name.

#AI #LLMs #MachineLearning
https://www.implicator.ai/karpathy-joins-anthropic-as-claude-moves-into-its-own-training-room/

I was just fooled by a deepfake video. 😱

I feel ashamed, frustrated, and annoyed — and worried that others will be misled too. The current warning system is grossly inadequate!

I started watching it and thought something was a little off because it seemed too perfect, but I ended up watching the whole thing anyway. It was recommended by YouTube’s algorithm.

I eventually asked Gemini (same mega-corp as YouTube) about it, and here is a central part of what it says:

“Ultimately, these channels survive because they pull in millions of views before the platform’s slower human-review and privacy-reporting systems catch up to them. Until platforms make AI disclosures massive and unmissable on the screen, these channels will keep fooling people.”

Full Gemini response:
https://g.co/gemini/share/18131e4379f6

#AI #DeepFake #YouTube #LLMs

WARNING: LLMs can generate convincing but entirely fabricated analyses of data. Two simple experiments show why AI-generated analysis shouldn't be trusted without verification.

https://www.conferencesthatwork.com/index.php/technology/2026/05/why-you-shouldnt-trust-llms-with-data-analysis

#LLMs #analysis #reliability #trust

RT @karpathy: Persönliches Update: Ich habe mich Anthropic angeschlossen. Ich glaube, dass die nächsten Jahre an der Frontlinie der LLMs (Large Language Models) besonders prägend sein werden. Ich freue mich sehr, dem Team hier beizutreten und wieder in die Forschung und Entwicklung einzusteigen. Meine Leidenschaft für Bildung bleibt ungebrochen, und ich plane, meine Arbeit in diesem Bereich zeitnah wieder aufzunehmen.

mehr auf Arint.info

#Anthropic #Bildung #Forschung #KI #LLMs #PersonalUpdate #arint_info

https://x.com/karpathy/status/2056753169888334312#m

"It’s all AI comments by fake people answered with fake replies by other fake people."

https://restofworld.org/2026/virtual-assistant-linkedin-engagement/

#news #linkedin #TechNews #technology #SocialMedia #AI #LLMs