Intelligence solves execution. It does not solve coordination.
AI makes bureaucracy cheap enough that you can accidentally generate infinite amounts of it.
The next scarcity is coherence. https://t.co/VXiilkjx25
Sam, can I ask how difficult the part of alignment was?
At 10,000-agent scale, you guys encountered a coordination shape in which local autonomous work had to be periodically distilled into shared useful state and redistributed, while humans remained an important control plane.
Do you think this is an issue worth solving?
This aligns so well to what I am working on. Not the Navier-Stokes solution (not sure what is that even).
But how they got there. How do you align 10,000 concurrent agents?
Agents decide meaning. Machines prove completeness.
This model represents a step-function improvement on many benchmarks, and its training is ongoing.
Our internal model group arrived at the Navier–Stokes solution in 88 hours, using around 10,000 coordinating AI agents.
Throughout the effort, we maintained the strict safeguards—including monitoring and isolation—that we apply to all our frontier evaluations.
Intelligence solves execution. It does not solve coordination.
AI makes bureaucracy cheap enough that you can accidentally generate infinite amounts of it.
The next scarcity is coherence. https://t.co/VXiilkjx25
@lennysan@bot@romanugarte_ #2, the point around unshipping was my favorite.
It's not only about what not to ship, but it's more able collapsing 'featurs' into the same UI.
One chat, any outcome. No need for more buttons, drop-down menus, etc.
I love this era of production.
Loved this idea: "The context of unshipping."
"There's this old school software tendency to say things like, "Groqbot now has..." But instead we can reframe it as, "Groqbot can now..." Which is, is I think a much more human way of kind of describing these capabilities. And I think it's forced us to think more in the frame of what are tools and, and what are capabilities that we can give Groqbot, not what are new things we can add to the product.
Making your bots reliably do really impactful work for you behind the scenes in a way that just works, and giving them the capabilities to do that, like, that's what users actually care about. Uh, and so I think in the context of unshipping, there have been a lot of Groqbot now has things, uh, that we've realized are actually just capabilities that don't need pixels. You know, let's kill as many pixels as we can."
Grok @Bot is the hottest new AI product in the world right now. I'm absolutely hooked.
It all started a few months ago as a tiny team holed up in a separate part of the office, building from scratch.
Four weeks after the first line of code, the whole company was obsessed. Three weeks later, it launched to the world. Less than a month in, millions of Grok Bots are already helping people get work done.
Roman Ugarte (@romanugarte_) incubated, launched, and now leads product for Grok Bot. This is the first time he's told the story:
🔸 Why they started fresh instead of building it into Cursor
🔸 The two non-obvious bets that made it such a success
🔸 What they learned from manually onboarding the first 200–300 users
🔸 Why they spent the weeks before launch *unshipping* features
🔸 The first prompt to try when you open it
Watch the full episode right here on X 👇
@lennysan@bot Both Roman and Tara from the previous episode say the future of work (and power tools) will be humans expressing intent, and then steering.
Love this.
@jetpacklouie Louie, scrubbing 100+ moments is the part I built https://t.co/Y90I6G7ifs for.
It watches the full video, identifies the best moments (how many you ask for), clips them, crops them, and posts them.
Use LUCIDREAM7 to get a 7 days free if you want to test it.
Anthony, I think https://t.co/McsXVmhaOG will help you with weekly clipping and posting for these platforms.
1. Upload/paste YT link
2. Get clips and post them.
Would love to know what you think, and here's a 7 days free trial voucher to you can run you first week. LUCIDREAM7.
We're spending a lot of time waiting for AI to complete its tasks.
I thought, why not make the wait beautiful.
This is how https://t.co/H4SfDRhOLa 'thinks'.
@eriktorenberg Very good reflection on this.
It took me 35 years to find my path, and two years on it I still didn't find success.
I can feel it's the path I was meant to take, but no evidence support that feeling.
The doubt is there, the "what might people think" is there.
It's all true.
The most important distinction Nevo makes here, which I really like is this:
- Replies to intent-filled posts capture existing demand.
- Outcome-led content creates demand among people who were not searching for software.