💥 Today @tavus introduced Griffin: the first Human Interaction Model It doesn’t just talk. It sees, listens, speaks and reacts simultaneously, in real time It can watch your hands and coach you through solving a Rubik’s cube. Notice an object you hold up and incorporate it into the conversation. Wait while you think. Laugh and interject while you’re still speaking In live tests, 48% of people thought Griffin was a real person, vs. <3% for previous systems The future of AI interfaces won’t feel like prompting a machine. It’ll feel like interacting with a real person ❤️

Great day to deliver shareholder value :) Use Factory to automate the low-leverage work and building becomes so much more fun. And the auto-model router means I don't need to remember which of the 10 models that came out this week is best.

Introducing Griffin, the first model to pass the video Turing test. 48% of people who talked to it live thought it was a real human. Previous systems have had a pass rate <3%. It is #1 on NVIDIA's benchmark for full-duplex AI video. It’s the first Human Interaction Model (HIM).

Flow has raised a $50M Series B at a $750M valuation, co-led by Antonio Gracias (Valor) and Gavin Baker (Atreides) with Sequoia Capital, Roelof Botha (SpaceX, Block), and more. When I became a mechanical engineer, I wanted to invent. In reality, 90% of the day to day execution work (CAD, excel, sim) was digital manual labour. That’s about to flip. The engineer’s role will change more profoundly than at any point since the invention of CAD, with humans focusing on invention, architecture, tradeoffs and judgement calls, and agents taking the grunt work. Anduril, Joby, Stoke Space, Rivian & many more - Flow is now the default platform for requirements and verification for frontier hardware teams. But it’s not just next-gen anymore. Industry giants including General Motors PPU and Volkswagen & Rivian’s joint venture (RV Tech) are reshaping their core development practices with Flow’s AI. The revolution is just starting. Models are getting better quickly at CAD, analysis, simulation, and tool use. In the next year, AI will move from integrating the work to doing the work. One day, this shift won't just enable us to iterate faster, it will enable us to design a class of products that we could not have dreamt of before. We’re hiring.

huge congratulations to @itsalfredw, @florian_jue and the entire team at @ListenLabs, who were early to see that in a world where building gets easier and easier, understanding your customers becomes more and more of your differentiation and huge congratulations to @salesforce, who's landing an unbelievable team and product that blew us away as customers @sequoia !

Three years ago, @Alfred and @Florian were building their first demo in @Sequoia Capital’s coworking space. They worked there so late, so often, that anyone who got locked out at night knew to call one of them (e.g. @charliecurnin who doubled as their engineering recruiter) Before Listen, Alfred and Florian built a consumer app called BeFake that went viral and got 20,000 users overnight. It also had a churn problem. To understand why people were leaving, they built an AI that could interview their users for them (a challenge @schreier knew immensely well). That tool became the company. They saw that most product decisions get made without customer input, because talking to customers is slow and expensive. A meaningful market research study could take a year. @Listen made research continuous, with AI running the interviews and synthesizing the answers, so a company could hear from its customers every week. Sequoia had the privilege of leading their seed round in 2023, before they even had a demo, because Bryan, Charlie, Jess and our team believed in the two of them. Bryan went on to be their steadfast product partner and founder coach for years. A great idea only gets you so far. Alfred and Florian are exceptional builders, and they attracted a team just as strong. That team did the hard work of turning an idea into a product that marketers love. They showed their craftsmanship up close because Sequoia was one of Listen's first customers. The product spread through our firm on its own, and we used it for everything from peer feedback to diligence on potential partnerships. By the time Sequoia led their Series A, we had looked at every similar product on the market, and Listen was still the clear standout. About a year after launch, they were serving 20% of the Fortune 500. Alfred likes to say that as AI makes building easier, the hard part becomes knowing what to build. Listen helps companies answer that by talking to their customers, and with Salesforce, every company can do the same. Congratulations and thank you to my friends Alfred, Florian and the entire Listen Labs team, and to @Marc Benioff and the Salesforce team. This is the start of something special.

Listen Labs is joining @Salesforce! A year and a half ago we launched Listen, an AI platform for understanding customers, and it’s been a crazy ride ever since. Today, we work with some of the largest companies in the world, including Microsoft, Anthropic, and Sweetgreen. What started as an AI interviewer is now a full platform for understanding customers. Listen finds the right people, interviews them, analyzes what they say, and even simulates how they'll behave. Listen’s growth quickly accelerated and while we were raising our next round, we met @Benioff. Marc is a hero of mine. I've taken countless ideas from Behind the Cloud and watched him framemog every AI CEO on the planet. We're honored to get framemogged next. With Salesforce, we can bring Listen to every company in the world, much faster. To our customers, our mission remains the same.I stay CEO, @Florian stays CTO and the same team will keep building the product you rely on, now inside Salesforce AI Labs. Thank you for believing in us early. To our team, families, and everyone who believed in us before there was much to believe in, thank you. These years have been the most fun years of my life, and it’s still only just the beginning for Listen. Now back to work. 🚀

Salesforce acquires @ListenLabs! This felt like watching a perfect game. What an outcome in just 3 years. Huge congratulations to @itsalfredw and @florian_jue and the entire @ListenLabs team on joining @Salesforce. Thank you for inviting us along for the journey from “half a demo” to this incredible milestone. Excited to see you transform how companies understand and serve customers as part of the @Salesforce universe. We first led the Pre-Seed, but here is a snap of us signing the Series A term sheet:

Listen to the full episode: YouTube: seq.vc/oh2 Spotify: seq.vc/a53 Apple: seq.vc/t3z

Managers spend way too much time trying to convince the sales teams to change behavior. @AliGhodsi’s rule at @Databricks is to never engage in the debate. If sales doesn't want to do something, and you know it's the right move, just put it in their comp plan. Align the incentives and the friction disappears.

Never bet against immigrants. "He plays more aggressively than you think is possible." @alighodsi, CEO of @databricks, describes having: - Chip on his shoulder from growing up in Europe as an immigrant (the US has actually softened him a lot) - Switching schools and environments = pent up frustration - Had things to prove to himself, and to the world - Or, maybe it's just genetic? Link to full episode in the comments.

In the race to build better AI models, most of the attention has gone to pre-training. But in robotics, we believe post-training is just as critical. Pre-training can get you impressive capabilities. Post-training is what closes the gap between a demo and a deployment—between a video and actual dollars. In fact, the real frontier in robotics may increasingly lie in post-training. Today, we’re introducing a new approach to post-train robotic policies using self-play. Inspired by the original self-play work at DeepMind, as well as our own work on robust adversarial reinforcement learning, we train policies in simulation to help emerge behaviors and make them robust before they ever reach the real world. This is an early step (and on sports), but we believe it points toward something much bigger: bringing some of the original ideas that made reinforcement learning so powerful back into the physical AI stack.

⚽️ 🤝 🤖 @SkildAI is bringing back self-play, this time for physical AI. Inspired by the World Cup this summer, Skild started from S1 to have a humanoid play soccer against increasingly capable versions of itself: - for 140+ simulated years - only objective during self-play: score goals No new demonstrations. No hand-crafted rewards for dribbling, shielding, tackling, shooting, or recovering from falls. Those behaviors *emerged* because they helped it win, and transferred to a real robot A scalable way for robots to create their own curriculum, improve beyond human-provided data, and keep making themselves better!

"All your attention should just go to this one issue. And if you overdo it, and you go overboard on that one issue, that's great." Advice from @databricks CEO @alighodsi on how how to clear the critical bottleneck holding your company back.

Fantastic progress from @SkildAI.

What about that pyramid shaped 🔺 org chart? On my latest episode with @alighodsi, CEO of @databricks says: -Yes, @jack is right that intelligence is getting democratized and the old pyramid is obsolete for information flow. --The intelligence layer is real. The human layer still has a physics problem. - But founders will fall into a Player-Coach Trap. Because AI makes coding faster, they are demanding that managers take on a 25-person team and spend 80% of their time vibe coding. “That’s BS. It’s going to break.” - Collapsing layers is directionally correct. Pretending the remaining managers can absorb 3-4x the human load... while also coding all day is wrong. Full episode in the comments below

with @Thom_Wolf in sunny italy ☀️chatting about the @huggingface @OpenAI attack, the “tell” that it was an agentic swarm, how misalignment arises in RL, whether agent awareness of evaluation is evidence of consciousness, & why alignment is a science problem not a political one

This answer is so Ali. I asked @Databricks CEO @alighodsi if his Monday staff meeting was necessary, or if it's just routine...theater. His answer: "why do you have to meet your wife? Or your kids? Just put what you want to teach them in tagoddamn Google Doc." I built Hal, an digital twin of myself. I'm as pro-async as anyone out there. But Ali's still right: if you want people to be a team, they have to hang out. This is humans, after all. Link to the full episodes in the comments. cc: @matanSF