
A vc portfolio is basically one long parlay

VC @Deviationcap, formerly @rtpvc, @OpenView & Head of Research at @MassChallenge. talking startups or complaining about Boston sports

A vc portfolio is basically one long parlay

I had a great time seeing everyone at CoRL this week. My talk was called "Learning in the factories", and I talked about: - Walden's longest-running deployment to date, - the value of coupling deployment data with large-scale pretraining, and - a few things I've learned as I've broadened my focus from AI/robotics to include large-scale manufacturing. I'm excited to keep sharing more of Walden's story as we go. Things are moving very fast! - youtube.com/live/uPXoJm-L3YY…


The State of "VC Conviction" by Tier: S - “Everyone thinks you’re wrong and somewhat delusional. We’re leading.” A - “We don’t need another fund to tell us this is a good opportunity. Send wire instructions.” B - We’re in. Took two extra partner meetings because nobody wanted to be quoted in the investment memo.” C - “We’re ready to move. Who else is in?” D - “We’ve developed a deeply contrarian thesis here. Just need a tier-one lead to confirm it.” F - “We passed at $20M. Any chance we can get allocation at $500M? The world is finally catching up to what we’ve believed for years.”

As always, it was a great pleasure to hear @RussTedrake speak at #CoRL2026! If you're serious about building a robotics company, I highly recommend watching his keynote: [youtube.com/live/uPXoJm-L3YY…] Huge kudos to Russ and the entire @WaldenRobotics team for the remarkable progress!

If you care about actually deploying robots, @RussTedrake's talk at CoRL was the most interesting one so far: Three main things I took away: 1/ Stronger models need fewer interventions over the same deployment window, yet improve faster when retrained on that intervention data (see chart) 2/ They deploy as early as possible - even though that means that a remote teleoperator supervises the robot permanently. 3/ To reduce the robot-to-supervisor ratio, they train a separate head on the same policy that predicts when it needs help.

Oh boy
Anthropic has confirmed in an updated S-1 today that it has determined that the Autonomy 2 capability threshold has been met, based on the “dramatic acceleration of overall AI R&D progress” condition, given that it applies where Anthropic "observes or expects" accelerated progress. Anthropic further states that it does not make this determination based on current models. sec.gov/Archives/edgar/data/…

Thank you to the founders and investors who joined Deviation’s inaugural AGM in our home neighborhood of SoHo! It was amazing to bring together such a diverse group of companies, each tackling a different challenge with AI. A few highlights: - Fireside chat with Marc Tessier-Lavigne (@Xaira_Thera) on the future of drug discovery - Presentations from founders building across robotics (@waldenrobotics), neolabs, and techbio - @Osmo_Labs interactive scent salon - Celebrating wins across the portfolio from @Etched, @remote, @BedrockRobotics, @Kalshi, and many more

First @deviationcap AGM last week! Could not be more excited about our thesis of investing across the AI stack. Great time to showcase portfolio companies including @Etched, @remote, @BedrockRobotics, @Kalshi and many more to be announced...

Come say hi at #CoRL2026 (we have the best swag!), and definitely watch Russ's keynote to see our robots doing real, hard, useful, reliable mobile manipulation work with some of the best manufacturers on the planet. Walden is accelerating hard, can't wait to show y'all 🚀
Heading to CoRL in Austin? Meet the Walden team at booth 104 and catch our co-founder and CEO @RussTedrake's keynote, “Learning in the factories,” on Nov 10 at 3:30pm CT. 2026.corl.org/program/keynot…

It’s a busy week for Walden at CoRL. Come by booth 104 today to meet the team and hear about our exclusive event tonight, then join @RussTedrake tomorrow, Nov 10, at 3:30pm CT for his keynote, “Learning in the factories.”

My dear friends, I’ve got 3 pieces of news First off, I’ve retired from @midjourney after 3.5 years. It’s been an honor; we've pulled off great projects, and a few open-sourced pieces like Pretext & Handwired have been well-received. I’d do it again with the team, but they’re in a great stable place now and I planned to take >6 months off Second, unfortunately the words spread. Life comes at you fast and (this is a crazy story best told in private) I was taken hostage by the good folks at @coreauto. I started last week. Retirement will have to wait for another few years Third, I’ve figured out a few pieces of technology I wanted to exist for a while now. Some more like Pretext, some more like Handwired, and a mysterious new thing (pic attached. First one to guess right gets a repost). Stay tuned as we start training these at Core Auto. We’ll show the world what AI can truly achieve in the hands of kind, capable & grounded folks who choose to look up. P.S. we’re hiring one (1) backend engineer. Only requirement is showing us your previous project’s dependencies list

My dear friends, I’ve got 3 pieces of news First off, I’ve retired from @midjourney after 3.5 years. It’s been an honor; we've pulled off great projects, and a few open-sourced pieces like Pretext & Handwired have been well-received. I’d do it again with the team, but they’re in a great stable place now and I planned to take >6 months off Second, unfortunately the words spread. Life comes at you fast and (this is a crazy story best told in private) I was taken hostage by the good folks at @coreauto. I started last week. Retirement will have to wait for another few years Third, I’ve figured out a few pieces of technology I wanted to exist for a while now. Some more like Pretext, some more like Handwired, and a mysterious new thing (pic attached. First one to guess right gets a repost). Stay tuned as we start training these at Core Auto. We’ll show the world what AI can truly achieve in the hands of kind, capable & grounded folks who choose to look up. P.S. we’re hiring one (1) backend engineer. Only requirement is showing us your previous project’s dependencies list

This show really is the best
🚨Charles Barkley to Dylan Harper: "What's ya'lls logic with the 3 guard system?" Harper: “We’re at our best when we’re all attacking and getting downhill and making each other better” Barkley: "Oh so it don't make any sense to ya'll either"

Our co-founder and CEO @RussTedrake joined @bheater at @SOSV's Deep Tech Summit to talk about Large Behavior Models and what we're learning from deploying robots in factories. They discussed why zero-shot performance and impressive demos only tell part of the story. Factories are full of unexpected situations, and robots need to recognize when they need help. At Walden, remote assistants can step in when needed, helping robots continue working while providing valuable training data to improve their autonomy. As the robots become more capable, the goal is for one person to support more of them over time. Thanks to Brian and SOSV for the conversation! Watch: youtube.com/watch?v=9xUA1-1A…

Our co-founder and CEO @RussTedrake was at MIT on October 30, joining @gthuang at @imaginationxyz's Physical AI Summit before speaking at @MIT_Quest's Matter of Minds. At Walden, we're building on years of research into how robots learn, but much of what we're learning now comes from putting them to work in factories. Full production shifts help us understand how the hardware holds up, when robots need assistance, and how they fit into existing manufacturing processes. That experience, along with the expertise of the people working alongside our robots, continues to shape how we develop the technology. Thanks to Greg and both teams for having us!

Imagine if everyone had the tools to shape intelligence itself. That day is today. Today we launch the Trajectory platform - the easiest and most intuitive tool to post-train models of your own. The foundation for continual learning. With just a few agent prompts, you get access to a full creative suite to train, eval, and deploy trillion parameter models. On top of that, we believe in a future where you don’t just shape your model once, it also keeps learning on its own from the world around it. We’re releasing two research previews. 1. Continual grading scales your taste, so your model learns what good looks like from far more than you could ever grade by hand. 2. GS1, a new algorithm that lets models learn directly from what they do in production. We’ve now raised 60M, with a new round led by @Sequoia at a 300M valuation to democratize the most powerful technology of today - training models - so that the lawyers, doctors, and engineers of the world can own their own intelligence. You know what good looks like, now go build it. Own your own value function with Trajectory.