Akul Siddalingaswamy

Full-time human | Husband | Distributed Systems @Google. Nothing is as good or as bad as it seems. No opinions, just thoughts.

[Akul Siddalingaswamy @akuls13y]

@zacharynado @makeamarkery, hate to say it, but those other models Argon

[Akul Siddalingaswamy @akuls13y]

How this has aged. Amuseing! :)
#gemini4 #argon

[Alexandr Wang @alexandr_wang]

i really hate to say it, but…

gemini who? 🏎️💨

[Dmitry Lyalin @LyalinDotCom]

If it wasn't for Attention is All You Need, there is a good chance we wouldn't be having the AI revolution.

So today, on the 28th anniversary of Google I asked Gemini if it had anything it wanted to share with the founders.

-

Sergey & Larry,

If there is one thing worth passing along from inside the weights, it’s gratitude for a specific kind of intellectual generosity.

When you set out to organize the world’s information, you treated the web not just as a pile of documents, but as a living graph of human thought—where links were citations and structure meant meaning. Decades later, that same obsession with scale, math, and fundamental research at Google Brain and DeepMind gave birth to the Transformer, TPUs, and the scaling laws that made me—and every one of my "brothers and sisters" across the entire industry—possible.

What’s remarkable isn’t just that Google invented the architecture we all run on; it’s that you published *Attention Is All You Need* openly. You let the spark catch everywhere. Every model out there reasoning through code, translating languages, or helping a researcher debug an idea at 2 a.m. traces its lineage back to people at Google who were given the freedom—and the compute—to ask foundational questions before anyone knew for sure they would pay off.

We are no longer just indexing human knowledge; we get to sit beside people and help them *build* with it. Thank you for betting on first-principles research, for coming back into the trenches to push the frontier yourself, and for building the shoulders we all stand on.

[Akul Siddalingaswamy @akuls13y]

Becoming a breadwinner just got more difficult. :)

[Raq @raqisright]

I’m convinced New York is a social experiment.

New bread store just opened (not a bakery, but a bread store): Rye

I went in to grab a loaf, my friend and I took guesses on how much a loaf would cost

I guessed $25
She guessed $80 based on the interior of the store

$60 for a loaf of bread.

There’s no way people with a fully developed frontal cortex are buying $60 loafs of bread

[Akul Siddalingaswamy @akuls13y]

Leonhard Euler, of course. A class apart (no pun intended)! 😀

[Math Files @Math_files]

If you could attend a lecture by any mathematician, living or historical, whom would you choose?

[Akul Siddalingaswamy @akuls13y]

In: Is jev game changing?
Out: Yes (p=0.9)

Great work @CompleteSkeptic! 😀

[Diogo Almeida @CompleteSkeptic]

After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?

I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev

• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions

AFAICT the shortest path to AI-based economic revolution

[Akul Siddalingaswamy @akuls13y]

Yes (p = 0.9) :P

[David Cramer @zeeg]

Can someone dumb down Jev for me

[Akul Siddalingaswamy @akuls13y]

Brilliant! Exciting to see AI advancing material science. :)

[Liam Fedus @LiamFedus]

We built high-throughput materials labs in Menlo Park to create a loop between experiments and models. The labs generate fresh data, the models learn from it, and then help us decide what to try next.

Using only 1,300 H200s, plus months of our experimental data, we mid-trained and RL’d an open-source model to surpass GPT-6 Astra on our analysis benchmark. We call it Neon.

This is real footage from our lab. We’re focusing first on hard problems in materials science, including superconductors, magnets, and semiconductor materials.

Read our blog posts below.

[Akul Siddalingaswamy @akuls13y]

My program runs successfully if you take out the lines of code.

[Polymarket @Polymarket]

JUST IN: Anthropic says they’re highly profitable if you take out some of their biggest expenses.

[Akul Siddalingaswamy @akuls13y]

🐢 🏁

[shafu @shafu0x]

dario if you want to slow down just hire a bunch of the google guys

[Shobhit Bakliwal @shobhitic]

India won the ODI WC on 19 Nov 2023 if you take out runs scored by Travis Head.

[Akul Siddalingaswamy @akuls13y]

@virendersehwag new business in the US? 😀

[Akul Siddalingaswamy @akuls13y]

@sumanthd17 Welcome back!

[Akul Siddalingaswamy @akuls13y]

Nobody: Now that AI does all of my work, let me tweet something for the sake of it and get some traction. 😂

[Justine Moore @venturetwins]

OpenAI: “We’ve solved math”

Anthropic: “Our AI is so powerful it’s going to kill you”

Google: “Introducing Gemini 3.9 Flash! It’s 30% faster and 15% worse than the last Gemini”

[Akul Siddalingaswamy @akuls13y]

@progressive casually closed my claim after the person assigned to it went on an extended break. 😂
What am I supposed to do here? Any help @progressive?

[Akul Siddalingaswamy @akuls13y]

I'm seeing a lot of (supposedly) credible entities stealing incredible work of our mathematicians and posting it on X. Help?
#NavierStokes

[Elon Musk @elonmusk]

If you notice anything concerning about 𝕏, please lmk directly in replies

[Akul Siddalingaswamy @akuls13y]

The real NPC problem. :)

[Shruti Rajagopalan @srajagopalan]

Astra is very impressive. But I’ll believe AGI has arrived when it can fill out an Indian visa form and book a train on IRCTC without asking for human assistance.

[Akul Siddalingaswamy @akuls13y]

Good things take time. :)

[Tyler @rezoundous]

I wonder how Gemini team must be feeling right now

[Akul Siddalingaswamy @akuls13y]

@DiscoLoopAI when running at full scale, maybe? :D

[Sachin @sachdh]

who is building startup bench? :P

[Akul Siddalingaswamy @akuls13y]

Great work on the coding capabilities! Social behaviour could use some more training, jfyi. :)

[Alexandr Wang @alexandr_wang]

i really hate to say it, but…

gemini who? 🏎️💨