Today we are announcing our biggest bet yet on our own silicon.
Starting next calendar year, we will scale our Maia family to multi-gigawatt deployments, making it the primary inference system for our own product and services.
And we’re introducing Maia 300, which delivers 40% more tokens per watt compared to leading GPU-based systems.
We are already deploying Maia 200 across the US, Europe, and Australia, delivering up to 1.7x more tokens per watt across experiences in Teams and Copilot.
Our goal is to build on the full-stack efficiency gains we’ve already seen, while continuing to bring the latest innovations from across the ecosystem into our fleet.
Dear President Elon,
Apologies for not realizing you’re the real boss of America.
Would you kindly allow BYD to compete with Tesla? Millions of Americans deserve affordable EVs.
Surely competition is good for everyone, unless it’s against you.
Thank you 🙏
INDIA IS DOOMED AS A COUNTRY
A professor who completed his education from USA and returned to teach at IIT Bombay was arrested for stopping a student from cheating.
This is the state of education in one of the most prestigious colleges.
The professor has resigned now. SHAME
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
Super proud of my repo file-ITR.
It has 170 starts, and 36 forks.
People with 8-10 years of industry experience are raising PRs on it and using it to streamline their ITR filling.
The reason I’m proud is because this is the first time I have done something which people are using.
Looking to connect with folks who actively work in 'legacy-ish' / niche stacks like COBOL, modern Fortran, Ada, Verilog, Erlang/Elixir, Perl.
If you’ve shipped real systems in any of these (or know someone who has), my DMs are open – want to route you to some interesting opportunities.
If you haven't, please 🔁 for good karma.
1/ Most people “prepare for interviews”.
If you’re aiming for top spots like @Rippling , @databricks , @stripe , @nvidia , @Razorpay , @JioHotstar etc, you may need to rebuild yourself for that bar.
Here’s the PLAYBOOK for 0–5 YOE folks in India.
Save this. 🔖
Been an honour to play a small part in this journey 👀🙂🎥
619 Million Views, 65.2 Million Peak Concurrency, handled…!
Looking forward to breaking more records and building deeper tech…
Yeah bro, just sprinkle ‘India’ on the sentence to become cool - zero research.
The global tech-education space is far more exploitative than most people imagine, but taking cheap shots at “India” somehow feels cooler.
Some quick facts:
- LeetCode did ~$33.7M last year. Hardly any videos, just by selling mostly company-tagged questions. Select “Google” on the list, and you will find “palindrome” 🤣 They smartly sell you pattern questions and you get a similar question. But it’s a foreign brand, so they must be right.
- Mock interviews cost $100–$300 per session.
- Salary-negotiation advice is sold for ~$100.
- Recorded AI courses go for $1,000–$2,000.
- Provoking ad’s “wanna be a swe at google”. Oops sorry, creative.
I am not even getting into the quality or customer support. It’s pathetic.
So yes, there are bad actors in the Indian market too, no one’s denying that. Some people have absolutely capitalised on insecurity, and they deserve to be called out.
But casually attaching a country’s name to look smart is not done here.
If you don’t want to pay to learn, that’s a personal choice. You can avoid it completely. No one forces you to.
People do pay for convenience. Almost every product you use today exists because of that.