overheard in silicon valley: "we're profitable if you don't factor in the token spend"

Quantum is poised to become a multi-trillion dollar industry in the next few years. We think New Yorkers should share in that profit. That's why we negotiated that New Yorkers get board seats and a 15% stake of intellectual property that comes out of NYU's Quantum Fabrication facility. Because all New Yorkers should benefit from the economic success of this city.

Is Muse Wonder full better than opus 5.5? I gave them the same prompt goal is to build me a triple aaa level lawn mower simulator i can drink beer and smoke a cigar. Api equivalent cost : muse - $4.11 opus - $99.43 It is very clear that muse is not anywhere near close to frontier in 3d tasks like video games, or even printable cad designs. Even scarier is it burned my $50 plan spending $4.11 in api equivalent tokens. wut?! i had to upgrade to $200 plan to finish this run.

Adobe Illustrator 2026 - How I Created a Realistic Corn Mockup

JUST IN: Association for Human Mathematics urges all mathematicians to boycott OpenAI following its release of 700+ mathematical research files.

behold, an evolutionary LISP kernel which uses llms to self-modify/hot-reload running application without downtime once all properties pass. just insert tokens, dreams and ambitions.

GET A BUSINESS PARTNER. if ur soloing business, ur leaving so much money on the table. find someone with your same level of obsession who can attack different departments of the business. instant 2x output. a great business partner is the biggest hack for growth.

also... i see why people hate astra...

More pre-training doesn't necessarily mean better generalization? This new paper discovers mode-hopping, where LLMs repeatedly switch between shallow pattern-matching and actual generalization, even while training loss remains stable. For example, OLMo3-32B went from 81% accuracy to 0%, then back to 81.7% within just 40B training tokens. On top of that, selecting an earlier 4.5T-token checkpoint instead of a 4.9T one improved GPQA transfer after math fine-tuning (36.3% vs 29.8%) and robustness to alignment attacks (53% vs 21%) alphaxiv.org/abs/2609.33150

Update: GPT 6 Astra Max reasoning Ultrafast: 6% weekly usage gone on $500 plan in 3 minutes just by asking it to install Rosetta 2 and open unity.

You are noticing that there is literally 0 differentiation in AI? Coding, tool use, computer use, personal agent, persistent memory etc. They also compete in price: We made this model, performs like that model, but 50% cheaper. Isn’t this straightforward commoditisation?

Hey Claude, Speedup everything