Adrián (@adrscott) header
Adrián (@adrscott) profile picture

thinking about minds, machines, and better futures AI/ML, @Livermore_Lab

Austin, TX
Joined May 2011
Photo by Adrián (@adrscott)Photo by Adrián (@adrscott)Photo by Adrián (@adrscott)Photo by Adrián (@adrscott)Photo by Adrián (@adrscott)Photo by Adrián (@adrscott)Photo by Adrián (@adrscott)Photo by Adrián (@adrscott)Photo by Adrián (@adrscott)Photo by Adrián (@adrscott)
My entire timeline is flooded with people preparing for some kind of doomsday scenario, yet nobody seems to know exactly what they’re preparing for 🤔 I bought a shit ton of water and canned food (mostly wild sardines, beans, and spam) just in case. Worst-case scenario, I’m prepared. Best-case scenario, I’ve got groceries for months 😅😅🙂
55
Adrián retweeted
Proprietary software had a good run.
67
104
14
1,790
86,459
A trillion dollars built by a guy who genuinely believed in things people thought were science fiction. Believe in your weird idea.
183
1,555
64
9,336
118,773
Adrián retweeted
Replying to @Pirat_Nation
Feels like we have forgotten games are meant to be enjoyed, not rushed. When Skyrim came out, I spent countless hours just wandering, getting lost in it. It was beautiful. This game is finally giving me that feeling again.
4
13
1
564
11,935
Adrián retweeted
"post-AGI, no one is going to work and the economy is going to collapse" "i am switching to polyphasic sleep because GPT-5.5 in codex is so good that i can't afford to be sleeping for such long stretches and miss out on working"
1,113
567
276
10,883
1,638,242
Adrián retweeted
Replying to @thsottiaux
At this point, I don’t really see why anyone would use Claude over Codex.
8
1
27
4,156
The longer you spend in tech, the stronger the urge to buy a farm and never touch a computer in your life again.
696
3,443
497
27,538
933,916
Excited to release new repo: nanochat! (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single, dependency-minimal codebase. You boot up a cloud GPU box, run a single script and in as little as 4 hours later you can talk to your own LLM in a ChatGPT-like web UI. It weighs ~8,000 lines of imo quite clean code to: - Train the tokenizer using a new Rust implementation - Pretrain a Transformer LLM on FineWeb, evaluate CORE score across a number of metrics - Midtrain on user-assistant conversations from SmolTalk, multiple choice questions, tool use. - SFT, evaluate the chat model on world knowledge multiple choice (ARC-E/C, MMLU), math (GSM8K), code (HumanEval) - RL the model optionally on GSM8K with "GRPO" - Efficient inference the model in an Engine with KV cache, simple prefill/decode, tool use (Python interpreter in a lightweight sandbox), talk to it over CLI or ChatGPT-like WebUI. - Write a single markdown report card, summarizing and gamifying the whole thing. Even for as low as ~$100 in cost (~4 hours on an 8XH100 node), you can train a little ChatGPT clone that you can kind of talk to, and which can write stories/poems, answer simple questions. About ~12 hours surpasses GPT-2 CORE metric. As you further scale up towards ~$1000 (~41.6 hours of training), it quickly becomes a lot more coherent and can solve simple math/code problems and take multiple choice tests. E.g. a depth 30 model trained for 24 hours (this is about equal to FLOPs of GPT-3 Small 125M and 1/1000th of GPT-3) gets into 40s on MMLU and 70s on ARC-Easy, 20s on GSM8K, etc. My goal is to get the full "strong baseline" stack into one cohesive, minimal, readable, hackable, maximally forkable repo. nanochat will be the capstone project of LLM101n (which is still being developed). I think it also has potential to grow into a research harness, or a benchmark, similar to nanoGPT before it. It is by no means finished, tuned or optimized (actually I think there's likely quite a bit of low-hanging fruit), but I think it's at a place where the overall skeleton is ok enough that it can go up on GitHub where all the parts of it can be improved. Link to repo and a detailed walkthrough of the nanochat speedrun is in the reply.
669
3,305
731
23,894
5,824,282
Adrián retweeted
i never took the dead internet theory that seriously but it seems like there are really a lot of LLM-run twitter accounts now
3,254
1,374
1,677
32,882
5,800,226
Men can stay up til 2 a.m., wake up at 6, be in debt, broke, alone, and still have faith that one day, everything will work out. It's called being a man.
1,377
29,246
1,360
222,918
7,941,170
Adrián retweeted
Replying to @vxunderground
Skill issue. Should've disguised it as a bug or feature instead of an obvious dead man's switch.
2
4
603
20,099
Adrián retweeted
me talking to an LLM >*swipe credit card for 200$* >*turn on max reasoning mode* >write "hello" >model thinks for 30 minutes >50 h100s light up >30 gallons of water consumed in cooling >3 children in africa die of thirst >"Hey! What do you need help with?"
123
637
50
13,099
431,071
🤔 *furiously starts DMing VCs*
Replying to @_mattwelter
Bootstrapping is a good idea. Good luck.
13
5,082
Bye Qwen3-235B-A22B, hello Qwen3-235B-A22B-2507! After talking with the community and thinking it through, we decided to stop using hybrid thinking mode. Instead, we’ll train Instruct and Thinking models separately so we can get the best quality possible. Today, we’re releasing Qwen3-235B-A22B-Instruct-2507 and its FP8 version for everyone. This model performs better than our last release, and we hope you’ll like it thanks to its strong overall abilities. Qwen Chat: chat.qwen.ai — just start chatting with the default model, and feel free to use the search button! HF: huggingface.co/Qwen/Qwen3-23… or huggingface.co/Qwen/Qwen3-23… ModelScope: modelscope.cn/models/Qwen/Qw… or modelscope.cn/models/Qwen/Qw… It’s smarter, knows more, can do more things, and works better on agent tasks. Try it out and see how it works for you. Check the model card on Hugging Face (huggingface.co/Qwen/Qwen3-23…) to see the benchmark results. This is a small update! Bigger things are coming soon!
200
535
278
3,703
995,990
That's what a man wants at the end
231
2,789
242
33,481
1,244,398
ChatGPT wants to replace Google. Claude wants to replace software engineers. Grok wants to replace your wife.
173
247
44
4,300
319,323
📚 If you’re a student choosing what to focus on, pick MATH. It will teach you to relentlessly rely on your own brain, think logically, break down problems, and solve them step by step in the right order. That’s the core skill you’ll need to build companies and manage projects.
972
4,254
465
34,741
1,671,058
If you want to reach your full potential and maintain clarity of mind, stay away from addictive substances. My success and health come from 20+ years of abstaining from alcohol, tobacco, coffee, pills, and illegal drugs. Short-term pleasure isn’t worth your future.
2,131
4,593
617
36,599
2,162,836
How we treated AI in 2023 vs 2025
522
3,045
447
29,532
1,121,737