Scott Wu explains how Cognition handled its first big enterprise customer. For a large bank, adopting new software usually takes 12 to 18 months: access to the code, security reviews, internal approvals. When Nubank signed, the whole Cognition team flew to Brazil. This is what they did there: 1. Went through every task the bank thought Devin could take on. 2. Did some of them by hand first, to understand what the work looked like. 3. Taught Devin to do each one correctly and built the orchestration around it. Scott says building the product was almost like building it for one company.

Cognition went from $1M to $1B ARR in 24 months. The launch demo got debunked. They had no customers. They cut the price from $500 to $20. Then they figured out what actually worked, bought a company 5x their size in 72 hours, and got their own AI agent writing 90% of their code. The full breakdown of how they built the enterprise machine, closed Goldman and the US Navy, and ran three funding rounds in 12 months: venturestreet.substack.com/p… This one is worth the read.

BREAKING: Cognition just showed every AI startup what enterprise sales actually looks like. Nubank signed with Cognition to use Devin, their AI software engineer. The entire Cognition team flew to Brazil to close it. For context: large banks normally take 12 to 18 months to adopt new software. Security reviews. Internal approvals. Code access. The full gauntlet. Cognition didn't wait for the process to come to them. They showed up. That's how you compress an 18-month sales cycle. Full breakdown here: productmarketfit.tech/p/the-…

The most funded AI agents right now can't book a multi-step trip. Instinct. Meta's Muse. Billions behind them. Both break the moment a task requires browsing, comparing, and deciding across steps. Chesky's read on why: the race for the 'main agent' seat is the wrong race entirely. Airbnb isn't building a chatbot. They're building something that bets on a completely different infrastructure future. The breakdown of who wins the agent era, and how Chesky is positioning against it: theaiopportunities.com/p/bri…

Most people measure success by output. Elon measures it differently. "The area under the curve of total utility is how useful you have been to your fellow human beings times how many people." That's a direct quote. And it's not inspiration — it's a formula. Impact = depth of value × breadth of reach. Which means a surgeon who saves 500 lives and a product that marginally improves daily life for 500 million people can occupy the same moral category. The math forces you to confront both variables simultaneously. Most ambitious people obsess over quality. They want to be excellent, to be the best in the room. But Elon's framing exposes a blind spot: excellence without scale is a rounding error on civilization's scoreboard. The hardest thing isn't being useful. It's being useful to a lot of people — which requires solving distribution, trust, and adoption, not just craft.

BIG MOVE: Elon Musk reportedly bought a domain to counter OpenAI's new agent product launch. OpenAI launched Dots, its new AI agent product. Musk's reported response wasn't a competing product. It was a domain acquisition — price and full details still unconfirmed. If true, the AI rivalry is playing out in some strange places. Is this a real strategic play or just noise?

Elon's hiring filter, in his own words: "If somebody executes well, I'm a huge fan. If they don't, I'm not." No pedigree. No politics. No charisma points. Pure output orientation. But the part that actually surprised me: he says he undervalued goodness of heart early on. Talent + drive + trustworthiness gets someone in the door. Character determines whether the org stays healthy long-term. Most founders learn this the hard way — hire a high-performer who poisons the culture, then spend 18 months cleaning it up. Elon's just saying it explicitly before you make that mistake.

Peter Thiel gave a 20-year-old Zuckerberg $500K for 10% of Facebook in 2004. Zuckerberg has said the money wasn't the valuable part. The sentence was: "In a world that's changing so quickly, the biggest risk you can take is not taking any risk." Most founders treat risk as the variable to minimize. Thiel's insight is the opposite: in fast-moving environments, caution compounds against you. Every month you don't move, someone else does. Every bet you don't take, the table shifts. The real danger isn't the bold move that fails. It's the safe move that costs you the window entirely.

Elon on superintelligence and jobs: "I think by far the most likely outcome is an age of abundance, where we don't have just universal basic income, we have universal high income." A reporter pushed back. What happens to the jobs? Elon's answer is a destination, not a transition plan. The optimist case has real logic behind it — superintelligence compresses costs, material scarcity collapses, abundance becomes table stakes. That's not a crazy projection. But "universal high income" and "no job" aren't the same problem. Income is economic. Work is identity, structure, purpose. Nobody's answered what replaces that architecture. The reporter's objection isn't Luddism. It's the harder question: what does human dignity look like when your labor isn't needed? Elon answered with abundance. That's not wrong. It's just incomplete.

BIG MOVE: Bill Gates said missing Android was his "greatest mistake ever." He pegged the cost at $400 billion in lost market cap. The backstory matters: in 1980 Microsoft promised IBM an OS it didn't have, then bought one for $75,000 to deliver. That improvised move built Windows into a dynasty. But mobile was a different race entirely. Android became the default OS for billions of devices. Microsoft never got close. Gates has been public about this regret before. The $400 billion number is his own framing of what the miss ultimately cost. Sometimes the biggest losses aren't bad bets. They're the ones you showed up too late for. Where exactly did Microsoft lose the plot on mobile?

BIG MOVE: Zuckerberg just told Trump he's spending $600 billion on AI through 2028. Not in an earnings call. Not in a press release. At a White House dinner, on a live mic, answering a direct question from the President. "Something like, I don't know, at least $600 billion through '28." At that scale, Meta isn't just competing in AI. It's trying to own the infrastructure layer entirely. No company has ever deployed capital at this speed in peacetime. Who gets crushed if Meta actually follows through?

THIS IS BIG: Dario Amodei is on record saying AI wipes out 50% of entry-level roles in law, consulting, and finance within 1 to 5 years. That's not a fringe prediction. That's the CEO of Anthropic putting a number on it. Entry-level is where every career starts. Kill the entry level and you hollow out the entire pipeline. No juniors means no partners, no MDs, no senior associates in a decade. The downstream effect isn't just job loss. It's the collapse of entire career ladders. The firms that figure out the 50% who survive will own the next era of professional services. Who gets hurt most if he's even half right?

Years before he became a billionaire, Mark Cuban got this from his business partner Todd Wagner: "You're a d*ck. People are afraid of you, and we don't get as much done because people are afraid of you." Most founders would get defensive. Cuban took it seriously. Fear-based management feels like leverage. It's actually drag. When people are afraid of you, they stop surfacing problems early. They stop pushing back on bad ideas. They optimize for not getting yelled at instead of getting things right. The best leaders don't make people afraid to fail. They make people afraid to stay silent.

BREAKING: Sam Altman just drew a direct line between AI and the concentration of power and wealth Elon said money won't matter by 2036. Altman's response: "Clearly I think money will still matter in 10 years." But the sharper line was this: his biggest fear with AI isn't misalignment or AGI risk. It's a small group of people ending up with all the money and all the control. That's the CEO of OpenAI saying the quiet part loud. When the builder of the most powerful AI system on earth names power concentration as his top concern, that's not PR. That's a signal. Who actually prevents that outcome if the people building it can't?

THIS NUMBER IS WILD: 98% of US households still aren't paying for AI. a16z just dropped their State of Markets II data and the penetration numbers are jarring. Paid AI subscriptions went from near zero in Jan 2023 to 2.2% of US households by April 2026. That's real growth. That's also 98% of the market untouched. Every founder fighting over early adopters is competing for 2% of the pie. The real opportunity hasn't even started. Who actually cracks mainstream AI monetization first?

Zuckerberg paid $1B for Instagram when it had 13 employees and zero revenue. Days before the deal closed, he had written internally: "Instagram can hurt us." That sentence is the whole thesis. This wasn't growth through acquisition — it was threat elimination dressed as expansion. In 2012, Facebook looked untouchable. But Zuckerberg saw what most CEOs miss: the thing that kills you doesn't look dangerous yet. It looks small, scrappy, and irrelevant to Wall Street. $1B for a 13-person app with no revenue sounds insane. Until you realize the alternative was competing against it at scale. The best strategic acquisitions aren't about buying what's already winning. They're about buying what could make you lose.

BREAKING: A 50-foot golden Prometheus statue just landed at Starbase. Cast in Paris. Crossed the Atlantic. Assembled in Texas. 100 days from nothing to finished. This is a French foundry planting a classical Greco-Roman titan at the edge of a rocket launch site. Prometheus stole fire from the gods and gave it to humanity. Someone is making the symbolism very obvious. The builders who think they're reshaping civilization are now commissioning the mythology to match. That is a very specific kind of confidence. Is this the most audacious piece of art in tech history, or a warning sign?

Elon Musk's actual risk philosophy is simpler than most people realize. "Failure is irrelevant unless it is catastrophic." Not every failed Falcon 9 test killed SpaceX. Not every near-bankruptcy moment killed Tesla. The ones that nearly did? Those got treated as existential. Everything else got treated as data. Most founders invert this. They catastrophize small setbacks and under-prepare for the ones that actually end companies. Know the difference between a wound and a fatal blow. Are most founders miscalibrating their fear?