I wrote Deep Learning with Python to be the definitive guide to how deep learning works and how to best make use of it. Tens of thousands of people got their career start via this book. 120,000 copies sold, and downloaded by millions more.
And now it's free to read online: https://t.co/3CbcQ7hmjp
Honestly, deployment friction might actually be the best thing that could happen to the AI infra sector. It forces the AI buildout to get stretched over a much longer timeline. That keeps pricing power strong for the companies above, and it gives the applications side of AI enough breathing room to demonstrate commercial viability and catch up to the datacenter capex. Avoids having a big capacity spike followed by a glut.
Either you believe the AI buildout is over, that AI was a flash in the pan, and that the need for compute has already peaked... or this a rare buying opportunity.
Reminder: GPU spot prices keep rising and it's hard to secure capacity. HBM and DRAM production is effectively sold out through 2029, with chip and memory supply projected to stay tight until well into the 2030s.
AI research isn't slowing down: frontier training compute is doubling every ~5 months, and model capabilities are scaling over twice as fast as they did pre-reasoning. Inference is profitable with 70-85% margins, and demand is growing geometrically. Each new model generation unlocks new use cases.
I'm not even an AI booster, just being realistic here.
Prices are rising because supply cannot keep up with demand. The above cos will keep growing 50%-100% per year and will be absolutely printing money for at least the next 5 years. That doesn't imply anything about share price, but I like to have ownership of unstoppable money printers. I'd buy the whole thing and take it private if I could
I believe there is a vast, extremely diverse space of possible approaches to learning (for both time scales: long-term plasticity / pretraining & on-the-fly adaptation / reasoning), and that the current stack (Transformer + gradient descent & token-space reasoning chain generation) is not even in the top 50% most efficient.
Future AI will look nothing like the current stack.
Flashback to June 14, 2024.
At the time, frontier AI models have zero fluid intelligence. They score around 5% on ARC 1. They can't reliably add two numbers. They make laughable logic errors every other prompt. Reasoning models don't exist. Coding agents don't exist. Claude Code doesn't exist. "Code gen" is a complete joke. Scaling up next-token prediction based on web text data is hitting a wall.
What if a time traveler told you that, by November 2026...
• AI models would have human-like fluid intelligence, thanks to a new paradigm: test-time induction (modern reasoning models).
• Not only would ARC 1 be saturated, but unreleased benchmarks 10x harder (ARC 2) and 100x harder (ARC 3) would be saturated too.
• Nearly all code would be written by AI, and models could one-shot complex codebases flawlessly.
• AI models would be solving multiple Millennium Prize problems.
• OpenAI's valuation would be 15x higher, and Anthropic's would be 85x higher.
And on top of all that...
• Nvidia would still hold ~80% of the AI compute market.
• Nvidia would have grown its revenue and earnings 4.5x.
• Nvidia would still be growing 75% a year.
• Nvidia would have even higher operating margins (67% vs. 65% back then).
Based on that information, straight from the future, would you have bought NVDA?
If you answered yes -- congrats. Today, 2.5 years later, you'd be down 5% on your investment (even more after inflation).
The big question of the moment is how to smoothly coordinate the work of many agents. One approach that is getting a lot of traction is to mirror human organizations and create agents with specific roles that act as researchers, PMs, testers, etc. Each agent maintains the context for its role.
They keep running in the cloud 24/7, and you can give them instructions from Slack, too. Worth checking out
𝐈𝐧𝐭𝐫𝐨𝐝𝐮𝐜𝐢𝐧𝐠 𝐌𝐲𝐂𝐥𝐚𝐰 𝟑.𝟎
Raised AI teammates with tools built in — not a blank agent you babysit.
•𝐀𝐠𝐞𝐧𝐭 𝐰𝐢𝐭𝐡 𝐚𝐧𝐲 𝐡𝐚𝐫𝐧𝐞𝐬𝐬 — Claude Code · Codex · Hermes · Pi · OpenClaw — shared context.
• 𝟓,𝟎𝟎𝟎+ 𝐥𝐢𝐯𝐞 𝐝𝐚𝐭𝐚 𝐬𝐨𝐮𝐫𝐜𝐞𝐬— Not a scavenger hunt for APIs.
• 𝐂𝐥𝐨𝐮𝐝 𝐬𝐮𝐩𝐞𝐫 𝐜𝐨𝐦𝐩𝐮𝐭𝐞𝐫 — up to 64 vCPU · 128GB RAM · 1.28T storage
• 𝐀𝐧𝐲 𝐜𝐡𝐚𝐧𝐧𝐞𝐥— Browser, Slack, Teams, Telegram, WhatsApp
Ready in ~30 seconds → https://t.co/fkimOApReb
BREAKING: A mathematician at Anthropic claims to have used the company's AI to solve the Riemann hypothesis, the most notorious open problem in all of math
https://t.co/nEXAyIkvSt
Announcing ⚡⚡
1) New records on ARC-AGI public eval !!
2) A new sample-efficiency benchmark
70% on ARC-1 with just 19M params (15% on ARC-2)
49% with just 1M params!!
(Latter beats TRM's 45% with ~400x fewer params)
Trained from scratch on only 1,307 ARC puzzles.
Just ~$13 to train
No text, no LLMs
🧵