From GPT-OSS system card:
First evidence of AI-driven optimization of LLM architecture.
"Hey O5, suggest 128 new SwiGLU variants and benchmark performance, doubling compute and halfing the number of variants at each step."
Very unpopular opinion:
I miss GPT 4.5 the most. Yes, it was slow, yes it has a weird attitude, but it was the only model that could actually write.
I'd be willing to pay a lot for access to a model like that.
Hope they release a model based on that pre-training run soon!
ML then vs now:
2015: invent new model, optimize for weeks on own hardware, win imagenet with ResNet: 75% accuracy
Total cost: $$$$
2025: Spin up 8 H100s in 10 mins, let Cursor write 100 lines of PyTorch, train for 2h: 76% accuracy
Total cost: $20
Thanks @PrimeIntellect
ML then vs now:
2015: invent new model, optimize for weeks on own hardware, win imagenet with ResNet: 75% accuracy
Total cost: $$$$
2025: Spin up 8 H100s in 10 mins, let Cursor write 100 lines of PyTorch, train for 2h: 76% accuracy
Total cost: $20
Thanks @PrimeIntellect https://t.co/9ivb5TChhk
With hyperscalers and neoclouds scaling GPU clusters into nirvana means compute actually got kinda cheap!
With Prime Intellect I can find the lowest rates at any given time and spin up a mini cluster in ~10mins.
$100 for 8 H100 for a full day 🤯
Somebody should hang this in a museum:
The original implementation of a ReLU neuron from the revolutionary AlexNet (2012) paper.
Copyright (c) 2011, Alex Krizhevsky
this is almost 95% similar to what ive experienced
its also something ive been trying to write about - I even have a chapter titled "do what feels right" trying to describe this (though honestly feel like below is more interesting)
this is almost 95% similar to what ive experienced
its also something ive been trying to write about - I even have a chapter titled "do what feels right" trying to describe this (though honestly feel like below is more interesting) https://t.co/Nkzpc9RCTv
14 Followers 58 Followingrebooting this handle for AI, engineering, tech, etc.
see @tomusually for personal shenanigans and https://t.co/r8ugK3IrZT for the professional stuff
357 Followers 317 FollowingTuring machine in a learning state optimized to try to build the impossible | GauntletAI Cohort 2 Grad | Wife | Not an AI bot.
136 Followers 40 FollowingPlatform powering the future of education.
Kids crush academics in 2 hours, get their time back, then dive into life skills and passions. 🚀🧠
5K Followers 2K Following“The only limit is the speed at which we learn.”-@sama | prev @arweaveeco | @DevRelUni Grad | @theNetworkState C1 | @joingauntletai S25 Grad | Prev Ranger Medic
561 Followers 1K Following(🇺🇸/cath/acc)—Citizen. Father. Data Center expert. Nuclear advocate. Investor in civilization. Ex-LinkedIn. (opinions my own, etc)
247 Followers 358 FollowingDevoted researcher on the topic of human advancement.
Language learning for busy people who want to master language and have fun: https://t.co/6S9nCAAjr6
67 Followers 217 FollowingFounder @ The Compute Index
AI Compute Infra @ Meta
Co-Founder @ Universe Energy
ML for options pricing @ UBS
AI ASIC chips @ ETHZurich
CPU Engineer @ ARM
704 Followers 8K Followingنحاول أن نعيش اللحظة، نضحك، نتسلى، نؤجل مواجهة الألم أو الانهيار، مع أننا في داخلنا نعرف أن هناك ثقلًا أو مصيرًا محتومًا ينتظرنا
8K Followers 650 FollowingAgentic Commerce: AI agents that aesthetically onboard each customer, collect contact info, and recommend products. Built for @Shopify.
Used by 4,000+ brands.
16K Followers 707 FollowingML Engineer @ML6team, part-time at @huggingface. @KU_Leuven grad. General interest in machine learning, deep learning. Making AI more accessible for everyone!
136 Followers 40 FollowingPlatform powering the future of education.
Kids crush academics in 2 hours, get their time back, then dive into life skills and passions. 🚀🧠
10K Followers 48 FollowingAn open-source declarative framework for building modular AI software. Programming—not prompting—LLMs via higher-level abstractions & optimizers.
5K Followers 2K Following“The only limit is the speed at which we learn.”-@sama | prev @arweaveeco | @DevRelUni Grad | @theNetworkState C1 | @joingauntletai S25 Grad | Prev Ranger Medic
62K Followers 12K FollowingAI policy researcher, wife guy in training, fan of cute animals and sci-fi, Substack writer, stealth-ish non-profit co-founder
648 Followers 1K Followingcurr @TradeADMIS, @METalumni assoc president 🌱 lets talk ai, energy, environmentals, and markets (all opinions here are my own)
7K Followers 576 FollowingOur global forests hold the secret to solving climate change and restoring our world's environment. We can be healthy, and our future can be bright! Join us!
306K Followers 1K FollowingI think agency might be the most important personality trait of the 21st century.
Read my essay 'High Agency' at https://t.co/3lfQgXXltI