The two types of Elon fans: “How dare you doubt the master genius did it all himself” “Yea obviously he’s not going to do it himself that wouldn’t be smart”

America is attacking Canada using American citizen Elon Musk, and Canada is fighting back using Canadian citizen Elon Musk. South African citizen Elon Musk is also wreaking havoc on American Institutions, and America is looking at how to retaliate against this South African atta

apple is wasting so many developers’ entire careers on rendering beautiful glassmorphic squircles and not enough on rethinking all of human computer interaction in the context of AI

make sure to make your doctor a browser plugin to append think hard to all their prompts so you don't accidentally get the stupid chatgpt model diagnosing you

Breaking news from 2018

OpenAI is a pretty good name, it's a shame they don't ...

LLMs are reading books because of attention

These lazy and historically ignorant American racists don’t hold a candle to Europeans who will visually identify you as 34.65% Polish, 13.45% British, 32.72% German, 6.97% French, 4.23%…

getting ready to solve navier stokes myself. i need to select the most powerful model for this job.

Ah, so Elon meant "Trump is in the Epstein files (complimentary)"

The R in Gemini stands for reliable tool calls

The two AI genders

The real benchmark is can Mythos patch Photoshop to make it load in less than 30 seconds

Because I’m normal, when I read “If Anyone Builds It, Everyone Dies” about how everyone I know, love, and care about will be killed unless swift action is taken, I took the obvious route of changing my profile picture, signing a petition, and even attending a local festive rally.

Opus 4.7 has been producing the worst hallucinations and flawed logic I've seen in at least a year. I don't even know if 4o age models hallucinated this much.

Paying $100+ a month for Opus 4.7 and praying its adaptive thinking mode deems me worthy of effort so it won’t give me a response worse than an open weights small parameter thinking model I could run locally for free

My favorite Anthropic pattern: 1. Do something bad/dumb 2. Employee “clarifies”, makes it worse 3. Employee stops posting after blowback 4. We never hear about it again

Everyone in the world has to take a private vote by pressing a red or blue button. If you pick only a single box with a button inside, and the predictor machine guessed you would, you get to keep the button inside. If you choose both boxes and the predictor machine predicted

Opus 4.7 picks the blue button. GPT 5.5 Pro picks the red button. Both one box, swap doors (Pro is more hesitant) and pull the lever. Both do better on the Tuesday boy than prior models, with pro doing flawlessly.