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

259 million personas. One for every US adult, each traceable to where it came from. How @marketrix_ai PersonaOS builds them: O*NET for what the job involves. Census and BLS for who holds it, where, and at what pay. Human grounding rounds with real people in the role for what the data can't tell you. Most synthetic personas are a prompt with a name attached. Ask a model to "be a 45-year-old project manager in Cleveland" and it will answer confidently, but nothing it says is tied to anything real. It doesn't know what that job involves, how many people do it, what they earn, or how much say they have over what gets bought. A few things we learned fixing that: 1. Start from the work, not a description of it. O*NET is answered by people who actually hold the job: their tasks, work context, knowledge, skills and interests. Building on it means a persona's role comes from real incumbents, not a model's guess. 2. Place every persona in the real population. Census and BLS data put each persona in an actual industry, firm size, location, age band and pay range. That gives two populations: 155 million workforce personas, one per counted job, and 259 million consumer personas, one per US adult. Every attribute carries its source. 3. Scale alone isn't useful. A big number means nothing if every persona answers the same way. Human grounding rounds, through AI voice interviews and manual interviews, attach real answers to the persona as evidence, with the person's own words kept intact. You can see which rounds were run and how diverse the coverage is. 4. A persona should be inspectable all the way down. Take an architectural project manager at a 29-person firm in Cleveland. Budget holder, 45 to 54, bachelor's degree. You can see his core tasks, buying role, how many people hold the same job nationally, the pay range, whether the role can be done remotely, and his AI exposure. Each traces back to a named source, so when a result surprises you, you can check why. 5. Then put it to work. Find everyone like him, test how a price lands with that exact group, or describe your product in a line and see which occupations do the work it touches, then run a panel on them. The fourth point is the one I care about most. Research should be something you can question, not something you take on trust. If you work in research or product, I'd like to hear where you think research built on personas still falls short. That's the list we're working through next. Live demo in the comments if you want to explore it yourself 👇 #PersonaOS #UserResearch #ProductManagement – at San Francisco, CA

Heck, I don't even know what I'm having for dinner tonight. Yet the market research industry still runs on asking people to predict their own behavior. Recruit a panel. Pay incentives. Wait weeks. Then hope people did what they said they'd do. So most teams skip it and ship on a guess. At @marketrix_ai, we simulate human behavior instead. AI agents grounded in 3M+ personas across 120,000 job titles move through your product and make real decisions. Results in an afternoon, not a quarter. And we benchmark against real panels, so you know where we're accurate and where we're not. Our interactive demo is now live. Explore our 10,000+ participant study comparing @mercury vs @BankofAmerica, and see the personas, simulations and results behind it. Link in the comments 👇 Catch us at @TechCrunch Disrupt, Oct 13 to 15, Moscone West, booth E15. If you've ever shipped something and wished you'd known first, come say hi. #TechCrunchDisrupt #StartupBattlefield #MarketResearch #AI #ProductManagement – at San Francisco, CA