AI agents are about to make millions of decisions for people. Not one of them knows how a person actually behaves.
That thought has been stuck in my head for a while. So I took it to the @TechCrunch Disrupt stage as part of the Startup Battlefield 200.
@yasithdev and I met in high school, building software for real businesses before we were old enough to know what a startup was. Every product we’ve shipped since has taught us the same thing: you never really know how people will react until it’s too late to change anything.
So at @marketrix_ai, we built a model of every adult in the US. 259 million personas, grounded in real data, so you can see how people will react to a product, a price or an agent’s decision before it goes live.
We call it the behavioral infrastructure layer for AI. Two minutes on stage felt like ten seconds.
Before your product meets a million real people, let it meet ours.
Your AI notetaker is blind.
My meetings are full of screen shares: charts, decks, dashboards. Every notetaker I tried gave me text notes or a pile of blurry screenshots.
So we built Nutmeg, the notetaker purpose-built for meetings with screen shares. It rebuilds what was shown into your notes as real charts, tables and numbers - then checks against what was said to fix mistakes and fill in details. No bots.
Your notes can finally see what you saw... and what you didn't! nutmeg.so
. @Apple nailed it with PROLED XDR. Best laptop display in the world, IMO.
The nano-texture lets me sit by a sunny window and read “quick updates” from 25 AI agents running in parallel.
Touch is useless and more of a distraction to me.
Cloudflare walked away from a $1 billion valuation. Six months later, it raised at less than half that.
That was the story I kept thinking about after hearing @Cloudflare co-founder @zatlyn in conversation with @FurqanR, who moderated the fireside at @fdotinc in Fort Mason. It was good to be back on campus. We spent five weeks there building @marketrix_ai as part of the Artifact batch.
I tried to put myself in their shoes. A billion-dollar valuation would’ve been easy to celebrate. Walking away meant living with that decision for another six months.
It left me with a question I’d want to ask in any fundraise. If we take this money, what will we be asked to give up?
Thanks, Michelle, for sharing the story, and @FurqanR, @hthieblot and the Founders, Inc. team for bringing us together.
Uber's CEO just told us what would earn a pilot.
Well... his digital twin did.
PersonaOS by @marketrix_ai builds digital twins of every U.S. adult (all 259 million) and of the executives you sell to, from the public record, with a source behind every fact. We also ground personas in real interviews: real people answer our AI interviewer, and the twins learn from what they say.
So we turned it on ourselves. The twins of ten CEOs reviewed @marketrix_ai: Marc Benioff (@Benioff, @salesforce), Dara Khosrowshahi (@dkhos, @Uber), Tony Xu (@t_xu, @DoorDash), Aaron Levie (@levie, @Box), Steve Huffman (@Reddit), Michelle Zatlyn (@zatlyn, @Cloudflare), Jensen Huang (@JensenHuang, @nvidia), Mark Zuckerberg (@finkd, @Meta), John Ternus (@johnternus, @Apple) and Eric Yuan (@ericsyuan, @Zoom).
1/ We pitched them
Each twin answered on its own.
→ 9 of 10 asked for the same proof: predict a test they've already run.
→ We offered one blind call. Not enough, said all ten. One right call could be luck. Call a batch, surprises included, and show the misses.
→ All 10 would start us on the same kind of flow: sign-up, onboarding or checkout.
2/ They used our live website
Each twin explored marketrix.ai alone, in a real browser.
→ 7 of 10 asked us to publish that proof, misses included.
→ 4 of 10 wanted to try it before signing up.
3/ We asked what they'd pay
Assuming Marketrix had passed the test they set:
→ 9 of 10 would sign a first-year platform fee of $100K–$250K.
→ Once it changed their decisions, 6 of 10 would renew at $250K or more. One went over $1M. Not one would drop it.
So our to-do list is clear: publish head-to-head results against real outcomes, misses included. Let people try before they sign up. Lead with sign-up, onboarding and checkout.
Three studies, ten CEOs, one honest to-do list, in under an hour. Every finding is a hypothesis until real people confirm it, and now we know which ones to test first.
The video replays those runs inside PersonaOS, condensed.
Try it: personaos.marketrix.ai
Every answer shown is simulated. Digital twins are built from public information and are never the person's own words. The people shown didn't take part in, review or endorse this video or Marketrix.
#ai#userresearch#productmanagement#b2bsales#digitaltwins
Clients have @claudeai now. So what is a law firm for?
@KhraminaDianka , Co-Founder and CEO of @runpraxis, backed by @ycombinator, says law firms rarely lose clients on legal work. They lose them on the experience.
It takes days to get a first reply and weeks of back and forth to onboard. Then comes a 100-email thread nobody can follow. In a business that runs on referrals, every bad experience costs a firm future clients.
Praxis is a system of record for law firms with AI agents on top that run the firm’s operations on its own rules. A lawyer approves every step.
Her take on where law is going is the part I keep thinking about. Lawyers are moving from doing the work to reviewing and approving it. Diana wrote her own company’s MSA with Claude and still had a lawyer check it, because what clients pay for now is trust and judgment.
We recorded it on @theteslapod while @Tesla FSD drove us through San Francisco.
We covered:
→ The moment a client walks away from a law firm
→ What a law firm is still worth when clients can ask Claude
→ Why she believes in a human in the loop, especially in law
→ Taking @gorgiasio's AI Agent from zero to 1,000 customers
→ Picking one idea out of 12 to work on for the next 10 years
No studio. No steering wheel. Just Tesla FSD.
Full episode in the comments.
Diana, thank you for joining me.
#theteslapod#teslafsd#ycombinator#legaltech#ai#startups
Great to be part of One Way Summit 2026 in San Francisco, bringing together founders, investors, and technology leaders to discuss the future of AI, entrepreneurship, and innovation.
The summit’s conversations connected AI infrastructure and its impact on society with the future of work, education, investment, and the immigrant founder journey.
Building the Infrastructure of Intelligence brought together @mlech26l (@liquidai), @zuzanna_pathway (@pathway_com), and @pilarmanchon (@Google), moderated by @readkrystalhu (@Reuters), to explore the foundations needed to scale AI and support its growing capabilities.
AI and the New Social Contract featured @annaadeola (@FoundationOAI), @satpugnet (@worldnetwork), and Martin Farach-Colton (@nyuniversity), with Lex Zhao (One Way Ventures), exploring how AI is reshaping our relationship with technology, institutions, and society.
@severinhacker (@duolingo), @iamcal (@SlackHQ), and @biilmann (@Netlify) explored how AI is changing the workplace, while @gaganbiyani (@theacademysf) and @katieschwartz (@a16z) tackled its impact on education.
On the investment side, @immad (@Mercury), @dessaigne (@ycombinator), and @SuzanneDiBianca (@salesforce) discussed emerging technologies and where venture investment could be heading in 2027.
And @AnnaPinol_ (@NFX), @anandc (@CelestaCapital), and @mananm (@UnshackledVC) explored the unique advantages and challenges immigrant founders face when raising capital.
What I find particularly interesting is how closely these themes connect - from the technology we build to the way we work, learn, invest, and create opportunities for founders.
It's great to see One Way Ventures creating a platform that brings these perspectives together while supporting immigrant entrepreneurs building companies in the US.
As an immigrant founder myself, that mission resonates with me.
Special thanks to @lex_zhao for inviting me and to the @onewayvc team for putting together a great summit. Really enjoyed the conversations and connecting with fellow founders.
#onewaysummit#onewayventures#ai#startups#immigrantfounders
This is Genie 4, our most realistic and interactive world model yet. 🌐
Explore dynamic environments in real time to see how the scene responds, sounds, and how the people in it move, react, and speak.
Did you know you can give your early suppliers a stake in your startup's upside, without handing over equity today?
It's called a warrant.
A warrant gives a partner the right to buy shares in your company at a set price, usually for a set number of years. They only benefit if you grow, and they're helping you grow.
Think about who takes the biggest risk on you early. A manufacturer taking your first order. A distributor giving you shelf space. A partner putting engineers on your integration. Most startups can't pay them what that risk is worth.
So instead of asking for a discount, you can offer them a stake in the outcome. The partners who believe in you early get rewarded for it.
I talked through this with @duane_wrnt, Co-Founder and CEO of Wrnt, on @theteslapod while the car drove us through San Francisco. Wrnt runs stock warrants from issuance to exercise, so companies stop tracking them in spreadsheets and rebuilding them at audit.
We covered:
→ What a warrant is, in plain English
→ How startups can use warrants to bring early suppliers and partners on board
→ What a warrant is worth to each side
→ What founders should know about dilution before they issue one
→ Why most warrant programs still get rebuilt at audit
If you're negotiating with a supplier or partner who's betting on you early, this one is worth your time.
Full episode in the comments.
Duane, thank you for joining me.
#theteslapod#startups#fintech#founderstories
Ben Horowitz's message to early-stage founders on betting on yourself, the doubts that do the most damage, and why fear means you're doing it right:
"Your job is to make your bet pay off. You believed in yourself enough to do it. There are gonna be many times when you question that bet and you question that belief, but that gets you nothing."
"The doubts are gonna be there. Ignore the doubts. The worst doubts aren't the ones people put on you. It's the ones that you put on yourself. Every entrepreneur has them."
"I spent a lot of time with Mark Zuckerberg when he was 22 years old and he had a tremendous doubt about whether he was gonna make it."
"It's not what you feel. It's what you do. You're gonna feel scared. If your guts aren't boiling, you're not even trying at this thing."
@bhorowitz w/ @speedrun GP @andrewchen
We asked 300 simulated small business owners about paying for ChatGPT for their team.
82% weren’t sure it would save enough time.
All 300 answered in about seven minutes. And 168 put their maximum willingness to pay at $11–$20 per seat.
That gives a team something specific to investigate: what would these buyers need to see to believe the subscription earns its place?
This is one of the studies in our new PersonaOS demo.
At @marketrix_ai, we’re building a way to explore how people might respond to your product, pricing, or message before you commit.
Start with a library of 259 million simulated personas. Describe the audience you want to understand. Then put a decision in front of them.
In this four-minute walkthrough:
→ Small business owners evaluate @OpenAI ChatGPT for their teams.
→ Restaurant owners compare @stripe and @Square’s homepages.
→ Simulated users navigate @Apple’s MacBook buying experience.
→ Surgeons consider Apple Vision Pro for surgical planning.
→ Truck drivers discuss the @Tesla Semi.
You can explore both the response and the reasoning behind it: where someone hesitates, what they misunderstand, and what makes them walk away.
Every result shown is simulated. Real people contribute answers through grounding, helping us refine how the personas respond.
A million users before your real users. This is what that looks like in practice.
Try PersonaOS: personaos.marketrix.ai
What would you test first?
#personaos#userresearch#productmanagement