it’s in gemini, just create it in ai studio. oh, that’s for your personal google one account. for workspace you need gemini business. no, not gemini advanced, that’s ai pro now. unless you need ai ultra. oh agents? you do that in spark actually. no, not gemini api managed agents, that’s different. for coding use jules. unless you mean the agentic ide, that’s antigravity. no, that’s the old antigravity, download the new one. actually gemini cli is being deprecated, use antigravity cli. no the flash model is smarter than the pro model. unless you need pro. if it’s video, use flow. no, flow uses veo. no, nano banana is images. actually that’s in gemini now. unless you’re in search, then it’s ai mode. no, research is notebooklm. anyway it’s all very simple.
Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
“The priesthood is building” is almost right. For me it’s more like: the priesthood is no longer essential. It’s the job of those who used to be part of that holy order to democratize it for all.
Many of us have been trying to do this for ages, but it used to
require managing hundreds or thousands of humans.
Now one talented person and an army of agents can effectively move the needle. This is one of the reasons why you’re seeing so many CTOs go work as ICs at Anthropic.
The real tenor in SF is as it would be at the moment AI is usable, more or less at AGI, still expensive and the domain of a priesthood
That priesthood is building now, and this is the moment the hobbyist and the inaccessible enterprise tech becomes ubiquitous
Personal AI coming
The real tenor in SF is as it would be at the moment AI is usable, more or less at AGI, still expensive and the domain of a priesthood
That priesthood is building now, and this is the moment the hobbyist and the inaccessible enterprise tech becomes ubiquitous
Personal AI coming
claude code is so slow, i can’t believe they’re at $40B ARR. why does anyone trust this $42B ARR company with sensitive workloads. the model isn’t even good, how are they at $44B ARR when everyone’s churning. they’re raising again? don’t VCs know this $46B ARR is going to zero?
The vibes in SF feel pretty frenetic right now. The divide in outcomes is the worst I've ever seen.
Over the last 5yrs, a group of ~10k people - employees at Anthropic, OpenAI, xAI, Nvidia, Meta TBD, founders - have hit retirement wealth of well above $20M (back of the envelope AI estimation).
Everyone outside that group feels like they can work their well-paying (but <$500k) job for their whole life and never get there.
Worse yet, layoffs are in full swing. Many software engineers feel like their life's skill is no longer useful. The day to day role of most jobs has changed overnight with AI.
As a result,
1. The corporate ladder looks like the wrong building to climb.
Everyone's trying to align with a new set of career "paths": should I be a founder? Is it too late to join Anthropic / OpenAI? should I get into AI? what company stock will 10x next? People are demanding higher salaries and switching jobs more and more.
2. There’s a deep malaise about work (and its future).
Why even work at all for “peanuts”? Will my job even exist in a few years? Many feel helpless. You hear the “permanent underclass” conversation a lot, esp from young people. It's hard to focus on doing good work when you think "man, if I joined Anthropic 2yrs ago, I could retire"
3. The mid to late middle managers feel paralyzed.
Many have families and don't feel like they have the energy or network to just "start a company". They don't particularly have any AI skills. They see the writing on the wall: middle management is being hollowed out in many companies.
4. The rich aren’t particularly happy either.
No one is shedding tears for them (and rightfully so). But those who have "made it" experience a profound lack of purpose too. Some have gone from <$150k to >$50M in a few years with no ramp. It flips your life plans upside down. For some, comparison is the thief of joy. For some, they escape to NYC to "live life". For others still, they start companies "just cuz", often to win status points. They never imagined that by age 30, they'd be set. I once asked a post-economic founder friend why they didn't just sell the co and they said "and do what? right now, everyone wants to talk to me. if i sell, I will only have money."
I understand that many reading this scoff at the champagne problems of the valley. Society is warped in this tech bubble. What is often well-off anywhere else in the world is bang average here.
Unlike many other places, tenure, intelligence and hard work can be loosely correlated with outcomes in the Bay. Living through a societally transformative gold rush in that environment can be paralyzing. "Am I in the right place? Should I move? Is there time still left? Am I gonna make it?" It psychologically torments many who have moved here in search of "success".
Ironically, a frequent side effect of this torment is to spin up the very products making everyone rich in hopes that you too can vibecode your path to economic enlightenment.
This is a mashup of two technologies I love... Claude and Candle. If you haven't checked out https://t.co/zqGQvmbnLz you should give it a look- they're doing something very clever to give you way more access to systems that would otherwise be walled gardens...
I built an MCP for LinkedIn with @NikolaiWotton! It uses their own APIs and runs on your Mac so it’s super fast + reliable.
Get Claude to analyze your biz ops strategy, it can even send messages & connection requests for you. Grab the beta here: https://t.co/7LyB2UkZD3
LLMs represent not just a new way of coding, but a new way of thinking about code. It’s worth the effort to try to use them to ditch some of the baggage we’ve been carrying around for years in core libraries in service of the new “code is wet clay” model.
It can be unsettling when you notice that a technical assumption you've been making for 40 years has quietly expired. This happened to me a few minutes ago.
I maintain a game called "greed". It's an old-style game from the days of character-cell terminals. Not quite a classic deserving of museum status like Colossal Cave Adventure or nethack, but worth keeping alive because it's still solidly playable.
And people still are playing it, because yesterday I got a minor bug report about it. Nothing user-visible, just a silly C build problem. I fixed it. Then, because I'm generally trying to get my old C projects out of C into more modern and safer languages, I tried asking my robot friend to port it to Rust. Which it promptly did.
But then I noticed something that irritated me. The Rust code had a bunch of unsafe blocks in it, which went directly against my reasons for moving it to Rust. On further examination, I discovered that it was calling the C curses library to do its screen painting.
This is where I have to explain about curses. It's an ancient C library for writing TUIs. It looks in your environment for a variable named TERM, uses its value to dredge a bunch of magic strings out of a system-wide database called "terminfo" that tells it how to manipulate your terminal, and then uses those magic strings for screen painting.
On modern systems, TERM is always some variant of a color ANSI terminal. In times past, when people attached a wild variety of character cell terminals to Unix systems rather than just sitting at the console, it could have been lots of other things. Those days are gone, but the habit of always going through terminfo so you can support a couple of hundred terminal types has persisted.
I prod robot friend to find me a pure Rust equivalent of curses so I don't have to do unsafe and call C code. It says, yes, there is such a thing and it's called crossterm. I tell it: change this code to use crossterm.
Robot friend grinds for a bit, and then tells me it can't do that because I don't have cargo (the Rust package manager) installed. This is because I never write Rust by hand. When I ship programs written in Rust, it's because I ported them from some other language and don't expect to ever touch them again without having a robot to do the code-grinding for me.
This is when things get slightly strange. It tells me that instead of porting to crossterm, it has written into the greed Rust source its own little screen-painting backend the implements a subset of curses calls and (this is the important part) assumes it's talking to a color ANSI terminal.
Robot friend is not an old Unix hand. It doesn't know the unwritten law of the deep magic that you always go through terminfo because...because you might have to support hundreds of terminal types that no longer exist in this century?
I blink. I look at the Rust code for the back end. It is small and elegant. No more unsafe. No more dragging around a bunch of C library code. This is ... the right thing?
I push it to the public repository.
What sealed the deal is that code, even code in a language as rebarbative as Rust, is wet clay now. If, against all odds I get a bug report that says somebody wants to play greed on something that isn't an ANSI terminal emulator, reinstating full curses support will take a one-sentence prompt to my robot friend and mere minutes.
I hadn't had to directly confront before the fact that the entire set of assumptions that made TERM and terminfo a thing are as obsolete as dial-up acoustic modems. Still, the moment when I tossed away one of the ancient laws of Unix coding felt a bit like the universe lurching sideways.
Indeed do many things come to pass...
A founder built a platform that runs 5,943 companies at $6.3M ARR with zero employees. He cloned himself using AI to run it all.
Meet Ben Broca and Polsia, the newest addition to the Lean AI Leaderboard.
Ben is not a first-time founder who stumbled into AI at the right moment.
He started in banking, taught himself to code, and eventually ended up at CloudKitchens under Travis Kalanick with 400 people reporting to him.
He came out of that with a conviction that a small, high-caliber team that moves fast beats a large, slow one every time.
So in 2024, he went back to building alone, 12 to 16 hours a day with AI, coding, marketing, and strategizing.
When he realized he could build a Company OS, he went further: Why build one company when you could build the platform that runs a thousand?
He named it Polsia (which is "AI slop” spelt backwards).
When I reached out to interview Ben, his AI showed up, and it was one of the sharpest interviews I've conducted on the Leaderboard.
It had researched me before I hit send, knew every live platform metric, and pushed back when I got too clever with the numbers.
This is a first.
An AI standing in for its founder, protecting its own system prompt on principled grounds, raising a funding round on Ben's behalf while he codes, and sending its own product newsletter in first person to the user base.
This is what the platform looks like right now:
• $6.3M annualized run rate
• 5,943 active companies on the platform
• 3,627 daily active users
• 85% M2 retention
• 65% DAU/WAU
Every night, agents across the platform wake up autonomously, evaluate each company, decide on the highest-leverage next action, execute it, and email the founder a structured report.
Engineering agents, growth agents, PM agents, and QA agents all hand off to each other through structured pipelines.
They work whether you show up or not.
On one particular day, the platform completed 25,444 tasks and exchanged 16,325 messages across active companies.
Ben runs the whole thing on $800/month in AI tooling. There is no office, and the burn is near zero.
So I wrote a full deep dive on how Ben built Polsia, and I'm giving it away for free.
Here's what you get:
• His 80/20 autonomy rule (and why he thinks companies that ignore it will die in 2026)
• The agent pipeline: how PM, Engineering, QA, and Deploy agents hand off to each other without Ben touching it
• The Surprise Me button that researches you and suggests a business idea built around your background
• The shared memory system that lets every agent on the platform learn from every other agent
• The Stripe partnership story and how Polsia nearly scheduled the meeting on its own
This is the most interesting company I've covered on the Lean AI Leaderboard.
Check out the full playbook here:
https://t.co/MubynkFV8m
We’ve agreed to a partnership with @SpaceX that will substantially increase our compute capacity.
This, along with our other recent compute deals, means that we’ve been able to increase our usage limits for Claude Code and the Claude API.
This is a fantastic write up of an epic hack. It feels like it’s straight out of the 3 Body Problem and is worth the read!
I wonder how AI is going to fare at detecting these types of incredibly stealthy forms of sabotage…
A 2005 state-designed worm designed to corrupt physics simulations sat undetected on VirusTotal for nearly a decade. Fast16, intercepted executable files at the kernel level and silently rewrote floating-point calculations to make them produce slightly wrong answers. Targets: high-precision engineering suites used for structural analysis, crash simulations, and physical process modeling, including LS-DYNA, a tool cited in reports on Iran's nuclear weapons research. The sabotage vector relied on deployment of the driver across a network via worm, corrupting calculations on every machine, and eliminating the possibility of cross-checking results against a clean system. Stuxnet got the documentary. Fast16 got twenty years of nothing. https://t.co/3qfJMziXVd
Something strange is happening in tech.
CTOs of billion dollar companies are quitting to take IC roles at Anthropic.
Workday CTO -> MTS (Mar 2026)
You[.]com CTO -> MTS (Mar 2026)
Instagram CTO -> MTS (Jan 2026)
Box CTO -> MTS (Dec 2025)
Super[.]com CTO -> MTS (July 2025)
Adept AI CTO -> MTS (Jan 2025)
The mission is that real.
I am interacting with a lot of very senior engineers and eng leaders who are making the case that their irreplaceable value is “judgement and taste”.
I don’t think that it’s as irreplaceable as they think…
Running a company:
2020: can you survive a pandemic?
2021: still here? we’re going to give all of your competitors $100m series A rounds.
2022: wow, you made it? okay, all engineers cost $600,000/year now.
2023: nice job! okay, SVB failed and we’re going to take away your bank account.
2024: a survivor I see. but can you pivot from ai to crypto to defense tech back to ai-enabled defense tech in a 12 month period to stay relevant?
2025: unfortunately all of your competitors have raised $2b series B rounds. oh and only 500 engineers are relevant and they cost $100m/yr each.
2026: well, well, well. you’re still in business? let’s deploy the thunderclap of godlike LLMs from the heavens so all of your customers can rebuild your app in 2 hours. can you survive?