Cobot

“There is no technology story that is not a human story.” Twenty-five years ago, I learned that in a class at MIT called the Structure of Engineering Revolutions. The professor was David Mindell. For our final project, my group went deep on Thinking Machines, the legendary supercomputer company. We had the circuit diagrams. What we didn’t have was why it really died. So we started cold-calling the people who were there. One of them, an architect at the center of the technical debate, broke down in tears on the phone still carrying the weight of a decision made decades earlier. That’s the moment I understood you can’t separate the machine from the people who built it. Today on Deployed, I finally get to turn the tables and interview my professor, David Mindell. MIT historian, deep-ocean roboticist, founder, and investor. Episode 6 is out now. Watch/listen here>> https://lnkd.in/grRHzQs8

When something is working, lean into it as hard as possible. That's the lesson Ross Fubini learned from 15 years of backing early-stage companies. And it's harder than it sounds. The instinct in venture is to diversify. Spread bets. Manage risk. But Ross watched the opposite pattern win, over and over. Anduril started as a collection of prototypes. Firefighting drones. Autonomous helicopters. Counter-UAV systems. Most of them didn't scale. But when the counter-drone product clicked when operators wanted it, when there was a mission mandate, when contracts started flowing, they didn't hedge. They went all in. More capital. More energy. More talent. Faster. The opposite pattern, the one that kills companies,  is the yellow light. Not in, not out, halfway committed. Leadership doesn't have conviction, so IT security slows things down. Partners don't want to hitch their wagon to something without sponsorship. I saw this constantly at Amazon. Over 60 robotics programs, most of them stuck in the mud because nobody would commit. The breakthrough came when we forced a decision: green light or red light, but no more yellow. That's the thread through this whole conversation. Whether it's robotics or defense tech or AI: when you find the thing that works, don't wait. Don't hedge. Go. Deployed Episode 4 is live. Link in comments.

There's a framework for selling products to the government that almost nobody outside of a handful of companies understands. It comes from Gokul, the CTO at Anduril. And Ross Fubini uses it to evaluate every defense company he backs. Three things have to be true: Operators have to want it. Not "this would be useful." Want. If you're not making a warfighter safer, a decision faster, an analyst more effective, with zero friction, you're dead. Done. Doesn't matter how good the demo is. There has to be a mission mandate. Someone in the system has to care enough to fight for this. "We're helping the DoD" isn't specific enough. Are you preparing for conflict in Asia? Responding to what we're learning about drone warfare? If the mission doesn't have urgency, the budget doesn't have priority. There has to be a contract vehicle. The government can't just write you a check because they like your product. There has to be a path, a SBIR, a Siver, an existing contract, where the money can actually flow. Miss any one of these and you're grinding. You might get a pilot. You might get a small contract. But you won't scale. The companies that break through, Palantir, Anduril, Scale, found all three. Usually after years of searching. This is the lens Ross uses to cut through the noise in defense tech. And it's the same lens that works for any complex enterprise sale: does the user want it, does someone have budget authority, is there a mechanism to buy it? Deployed Episode 4 is live. Link in comments.

Does OpenAI go to zero the way Netscape did? That's the question Ross Fubini asked me on Deployed. It’s not a clean answer. We both lived through Netscape. We watched the browser wars. We watched Yahoo and Excite and AltaVista disappear. We watched a company that defined the internet become a footnote. We watched the birth of social media and big data. We watched the future unfold. Ross's take: the “homepage” for AI hasn't shown up yet. We're in a discovery phase. Somewhere in the current wave of AI companies is the equivalent of the proxy server team at Netscape a side project that nobody thought mattered, building what would eventually become the CDN industry. Billions of dollars. From a team nobody remembers. And somewhere else is the calendar server team. Fully funded. Great engineers. Building something that would never matter. The difference between those outcomes isn't technology. It's execution. It's figuring out where the value actually accretes. It's not getting enamored by the demo. That's the same lesson Ross learned backing defense companies for 15 years. The best technology doesn't always win. The teams that find all three pieces: operators, mandate, contract, are the ones that break through. We also talk about the loneliest part of being a CEO, why patience is the scarcest resource in venture, and what we've learned about each other over 30 years. Episode 4 is live. Watch now: https://lnkd.in/gSSh8VNY

Many years ago, two nerdy kids showed up to interview for internship at Netscape on the same day in Mountain View. One came from a 250-person farm town in upstate New York. The other came from Northern Virginia, grandson of an Assistant Secretary of Defense, but had never seen anything like Silicon Valley. Neither had a master plan. They just kept walking toward the hardest, most interesting thing they could find. One became a Distinguished Engineer at Amazon, then built the robotics org, then started Cobot. The other introduced the first two employees at a company called Palantir. That introduction might be the highest-value intro ever made. Then he spent 15 years backing 40 more companies from that network, including the first check into Anduril. Tomorrow on Deployed, Ross Fubini and I talk about what we've learned since those cubicles at Netscape. Why the best opportunities don't announce themselves. Why execution matters more than magic. And why the defense tech boom has a talent problem nobody's talking about. Episode 4 drops tomorrow.

Remember the Sony robot dog, AIBO? For its time, it was genuinely magical. A robot dog that could walk, respond, learn. The technology was real. The engineering was impressive. The people building it believed they were changing the world. But as Bilal Zuberi put it on Deployed, will enough people buy it? What tricks does it need to do for you to stay engaged with it? Nobody knew. And eventually, Sony killed it. That's the gap most deep tech founders fall into. They solve the science problem and assume the business will follow. It almost never does. A product needs customers. Customers need to pay you. Revenue needs margin. That chain sounds like a day-one business school lesson. But most of the pitches Bilal sees can't connect all three links. The founders who survive deep tech aren't necessarily the ones with the best technology. They're the ones who understand that proving the physics is step one, and step one is maybe 20% of the job. The other 80% is figuring out whether anyone will write you a check, repeatedly, with enough margin for you to do it again tomorrow. We got into this and a lot more in Episode 3 of Deployed. Link in comments.