There's a reason @cognition keeps winning deals against companies 10x their size. @vvkgopalan sat down with @theodormarcu, @bryceehunt and @moritz_stephan (all former founders!) to discuss hiring philosophy, infinity stories, enterprise coding agents, and what it actually takes to compete. (00:00) Introduction (03:45) Cognition team & hiring philosophy (06:00) Infinity stories (13:15) Product surface area and future goals (17:20) Activation energy, experimentation, and customer-driven innovation (19:00) Adoption, operational changes, and ownership (22:05) Vibe coding hot takes (24:15) AI-first engineering organizations and managing AI agents (28:40) Future of software engineering roles and misconceptions about Devin (30:00) Model development, learnings, and Cognition's journey in vertical integration (38:10) What should you do as a new grad?

Working with close friends is one of the rare privileges this job gives you. Watching them go from idea to a real ~business~ is an even bigger one. Proud to call these guys friends. Excited for everyone to finally see what they've been building for the last two years.

Two US aviators went down near the Strait of Hormuz. DoW used an autonomous @Saronic Corsair to find and bring them home - the 1st drone rescue at sea in military history. This is why we build: to keep our troops safe. This was science fiction. American builders made it real 🇺🇸.

In the Summer of 2022 @MavrookasD asked @8vc's design team to mock up scenes of what it would look like to have an autonomous vessel rescue a warfighter in distress. Almost exactly 4 years later @Saronic is worth 9.25bn and is rescuing warfighters in distress. 🚀

Generalist CEO @peteflorence says robotics models are in a transition period similar to the step change between GPT-2 and GPT-3. They're "starting to cross over into levels of performance where these things are commercially viable for a number of different applications." "We think this is a crossover point where we have a general model starting to be able to hit levels of reliability, speed, and improvisational intelligence where we can start to get these things out there." "Very much like — you take a GPT-2-level model, you scale it to a GPT-3-level model, and certain types of commercial applications start to become viable."

AI is creating problems it still can’t solve. The same technology poised to automate millions of jobs still can’t reliably help people navigate SNAP — the food assistance program 40 million Americans depend on. We built the first benchmark to measure that. Partnering with Center for Civic Futures and @codeforamerica , we scored models on SNAP question scenarios which users would have to navigate, with expected response rubrics validated by policy experts to match practice considerations. The best model only scored 62%. Models handle federal questions like appeals and recertification reasonably well, but fall short on state-specific ones like replacing an EBT card. Benefits are administered by the states, meaning models are weakest where people need them most. The same technology poised to automate millions of jobs should at least help strengthen the social safety net for the people it could displace. As AI adoption proliferates into public services, governments need a reliable way to test these tools before deploying them.

In the 1980s, the US produced 80% of global enriched uranium. Today that number is zero -- and we rely on Russia instead! @ScottNolan & @GeneralMatter are solving this crisis; he’s jumpstarting America's nuclear renaissance, and makes us all more bullish on its scale & speed.

Robotics is going to get its Claude Code moment long before its ChatGPT moment. Everyone keeps waiting on the big general-purpose humanoid demo, the one that's supposed to make the whole thing click. But that's not how LLMs broke through. Coding got there first. It was narrow, hard, and mostly ignored. But it was the first place models got good enough to pay for, and that usage and data loop compounded into everything people point to now. Robotics has the same kind of wedge in front of it: dexterity. The precise manipulation that real work is made of, like plugging a cable into a socket, kitting, repackaging and assembly. It's the hardest unsolved problem in the field, and the first thing that actually turns into revenue. Whoever cracks that owns most of what comes after. Which is why we backed @GeneralistAI's $400M round. More here: 8vc.com/resources/announcing…

It takes a village, over half a billion in total funding, a lot of compute, and an incredible team to build general intelligence for the physical world. Join us: generalistai.com/careers

Ten years ago I set out to find the team building general robotic intelligence, and @GeneralistAI is it Their rigor, approach, and work on scaling laws give them the best shot at resolving Moravec's paradox this decade Why we love them: 8vc.com/resources/announcing…

Today, Ramp raised $750M at a $44B valuation. Last time we grew this fast, we were 1/20th the size. For 2000 years, business was built on two pillars. Today, a third: intelligence. It’s your least governed cost. It’s also your single greatest opportunity.

You might have access to a cardiologist or dermatologist, but many Americans do not. In fact >50% ofthe counties in the US don’t have a single practicing cardiologist. This is in part why @martinvars founded Certuma and why we are proud to have led the seed round at @8vc. This article in WaPo hits upon many of the drawbacks and concerns around introducing clinical AI. I think the article leans too much in to a job replacement narrative that does not seem viable or likely (medicine is about more than eg the discrete task of prescribing). And as to what happens when it makes a mistake? AI is readily auditable and its performance can be measured more directly and frequently in comparison to a doctor or nurse. Is it believable that we could provably improve these systems to a point of achieving FDA approval?

Our design team is incredible. Custom microsite for our Rocket League tournament this Friday with @cognition. Melee and other games will be featured as well. @ScottWu46 will probably school everyone, but that shouldn't stop you from trying. Limited spots @ rocketdevin.ai

Devin Desktop is our first product launch that’s fully agent-neutral. You can run your own custom background agents directly from the desktop app, Devin, or even Claude Code / Codex. Part of Cognition being the Independent Agent Lab is working well with all of the agents - excited to do more in this direction.

NEW: @JTLonsdale with @MariaBartiromo on @cognition's latest $1B round "What's really fun about this AI wave is the returns have really gone to the most talented -- physics Olympiad people, the best programmers in the world... The CEO @ScottWu46 used to work for us at Addepar a long time ago after winning the global championship in programing. He's brought together a bunch of other gold medalists... Cognition is going into big companies, like the big banks here in New York. They're working with them and saying, 'This project would have taken you four years... Let's figure out how you guys are going to use our agents... and we'll get it done in 3 or 6 months instead.'"

"This is what it would have been like to be friends with Thomas Edison." That's how Michael Thompson describes our founder @joeben, and it's why he helped take Joby public in 2021 through Reinvent Technology Partners. When Michael diligences a new technology in a highly regulated industry, his non-negotiable is an inventive, technical founder at the helm. 🎙️ Hear the full thesis on Joby Recon Ep. 04 "Backing the Business" hosted by @maxabusa. Out this week.

Excited to quadruple down here! @ScottWu46 and @russelljkaplan have been amazing partners to us and remarkable thought leaders in Agentic AI

Big milestone today. But still very early for self-driving software. When I was working on Autopilot at Tesla, we had an internal infra project called "Operation Vacation" - the goal being that the entire Autopilot engineering team could take a vacation and come back to a better self-driving stack as long as the data kept flowing and the models kept training. We'll soon see the same idea come to life for entire software products, managed and improved by fleets of Devins. Last year, our usage graph had a big dip during the holidays. A goal on our engineering team is that the same time this year has no dip, because the Devins are proactive and working by default. Automations are live in the product already - give it a try!

Grateful for our customers, investors, partners, and most of all for our incredible team. Hard to believe it's only been two years. Still a long way to go to build the future of software engineering - we are hard at work every day on making Devin even better! If you haven’t tried it in a bit, check it out and let us know what you think: devin.ai