Operating Partner @Unusual_VC Former @AppDynamics @Amazon, @Holloway Contributor @uarizona (opinions are my own)
San Francisco, CA
Joined April 2009
- Tweets5.5K
- Following1.7K
- Followers526
- Likes2.4K
Pinned Tweet
As an early stage founder, at every point during the recruiting process, you have to be selling candidates on why your startup is going to be the next great opportunity in their career. bit.ly/30YIvDK
Jon retweeted
We just released an Unusual Guide: “The New Physics of Go-To-Market in an AI-First World” by @jaredwaxman, with a foreword by @jvrionis.
Pricing, hiring, outbound, AEO and more—drawing on lessons from the Unusual Ventures portfolio. If you lead growth or GTM after product-market fit, this is for you. gtm.unusual.vc/
Jon retweeted
Excited to welcome Dream Labs to the Unusual Ventures portfolio!
Congratulations to the team on launching Dream-1, a big step toward robots that adapt to new tasks and environments. Onward!
Today, Dream Labs is coming out of stealth with Dream-1, a robot foundation model that controls bimanual robots it has never seen, on tasks it was never shown, in places it has never been, from $3/hour.
Dream API is live for everyone, today. 🧵
Excited to welcome @dreamlabs_ai to the @Unusual_VC portfolio
Today, Dream Labs is out of stealth.
In February 2026, while we were at NVIDIA, @scott_e_reed @youliangtan @GeYunhao and I started calling a new kind of robot model a World Action Model, one that learns how the world moves from video, then uses that to decide how to act.
Today, we’re releasing Dream-1 through a public API that anyone in the world can use.
When you rent a car you’ve never driven, you don’t learn to drive all over again. Within a few minutes, it feels like your own. Robots should work the same way: a good model should be able to pick up a new robot and get to work, without being retrained from scratch.
That’s what we’ve spent the last 6 months on, and it’s what Dream-1 does. It controls bimanual robots with parallel grippers it has never seen before, zero-shot, on unseen tasks in unseen environments.
Generalization has been my North Star in robotics: first by distilling video world models into VLAs (DreamGen), then by learning to act from those world models directly (DreamZero), and now by transferring across robots (Dream-1). I’ve come to believe it’s the only direction that lasts. Pushing a narrow model from 90% to 99.9% on a single task is real work, but much of it gets washed away once a model that generalizes arrives.
Since the start of the company we’ve stayed single-minded about generalization.That focus is what got Dream-1 to control robots it was never trained on, on tasks it was never shown. And it’s what makes a public API possible at all: a model that only does one task on one robot can’t serve millions of users.
Dream-1 is still a research beta, and it will make mistakes. But it already drives cars it’s never seen.
Generalization wins in the end, and the end is closer than it looks.
Here’s Dream-1 taking the wheel of a robot it’s never seen. Full details 👇
Jon retweeted
Optimizing local inference for ultra-low latency is both easier and harder than optimizing for a cloud service.
The unavoidable hardware variety, memory constraints, and thermal throttling automatically make each optimization more complicated.
But knowing which challenges to tackle first is easier because you’re dealing with users who:
1) know exactly what they’re looking for from inference performance for their setup and
2) are predisposed to build in the open and share their findings
Here’s how we know: @JiaZhihao announced that we were open-sourcing lithos-metal, our local inference engine for Apple silicon, 5 days ago.
In that time, we’ve made a ton of progress because everyone who let us know they tried lithos-metal (or even just explored the code or blog) gave us the platonic ideal of technical feedback: immediate, specific, actionable.
Thanks to all of you who shared your questions and suggestions, including insights into your actual use cases and testing approach. Our day job is still primarily building ultra low-latency agent infrastructure on standard GPUs, but it’s important to us to keep lithos-metal going and build the community.
Internally, we talk a lot about taking a co-design approach to AI efficiency. That means simultaneously thinking about how the hardware and software layers impact each other. Only focusing on one area for improvement leaves performance gains on the table.
That logic applies to where your inference engine lives, too. Whether it’s in the cloud, on-prem, or on your laptop, we’re excited to keep working to solve agent latency for our users.
🤖 Made with AI
Great insight from @Henry_Buck_
Great launch from @LithosAI: a free, open-source tool that makes AI run much faster right on your laptop. Mac users, give it a try 👇
lithos-metal runs fast... like Qwen3.8-27B at 200 tok/s peak on an M5 Max chip fast 🚀 And now it's open-source.
The LithosAI team combined megakernel and DSpark optimizations for @Apple silicon: lithos-metal uses one Metal megakernel per layer mixer.
That's how it runs Qwen3.8-27B 2.1x faster than MLX, 2.4x faster than Ollama, 4.5x faster than vLLM-Metal at 32K context.
Explore the code, try it on your Mac, and share what you learn. We'll keep working to speed up local inference across more models and devices.
Great launch from @LithosAI: a free, open-source tool that makes AI run much faster right on your laptop. Mac users, give it a try 👇
We’re open-sourcing lithos-metal 🚀
Megakernels + DSpark speculative decoding run Qwen3.8-27B at 200+ tokens/s/user peak on one @Apple M5 Max.
Ultra-fast inference on your laptop. Try it with any coding agent in one command.
Code: github.com/lithos-ai/lithos-…
Tech blog: lithosai.com/blog/lithos-met…
Check this out and let @thatniamhod if youre interested
Hosting a private dinner in SF with @nclando and @dimitriosSkar in the coming weeks on “The Speed of Intelligence: What Faster Inference Unlocks for AI Agents,” with engineering leaders from @DoorDash, @Uber, @augmentcode and more. If interested, HMU.
Jon retweeted
Hosting a private dinner in SF with @nclando and @dimitriosSkar in the coming weeks on “The Speed of Intelligence: What Faster Inference Unlocks for AI Agents,” with engineering leaders from @DoorDash, @Uber, @augmentcode and more. If interested, HMU.
Jon retweeted
Proud of the @MiravoiceAI team for reaching this important milestone!
We’re excited to share that @MiravoiceAI is now SOC 2 Type I compliant!
This is an important step towards building confidence in the practices that we use to keep customer data safe and secure.
🧵1/3
Jon retweeted
Super excited to release LithosBox, our ultra-fast sandboxes for AI agents! 🚀
17× faster startup, 43× faster snapshots, 16× faster branching, and 6× faster restores than existing sandbox providers.
🎉 FREE for the next two weeks during public preview. Give it a try!
Today, LithosAI is launching a public preview of LithosBox: ultra fast versioned sandboxes that your agents can snapshot, fork, and restore in milliseconds.
So far, we've been focused on providing ultra-low latency inference for open models like Kimi K3, DeepSeek V4.1, and GLM 5.3 Flash (stay tuned!). Now, we're giving agents runtime and post-training environments with Git-like control.
With LithosBox, agents can:
→ spin up, branch, and snapshot thousands of environments as they explore multiple paths forward
→ decide on a winner and rebuild the winning state
→ then throw away the environments they don't need, all in milliseconds
@dimitriosSkar, @JiaZhihao, and the rest of our team put together this agent vs. agent showdown with Doom so you can see how LithosAI helps your agents run 5x faster with AI serving infrastructure co-designed for ultra-low latency across inference and execution environments.
In our demo, the agent completes its entire run 5.29x faster with LithosBox and Lithos DeepSeek V4.1 Flash inference than it does with leading sandbox and inference providers.
Try LithosBox for free for up to two weeks via our public API.
Jon retweeted
Today, LithosAI is launching a public preview of LithosBox: ultra fast versioned sandboxes that your agents can snapshot, fork, and restore in milliseconds.
So far, we've been focused on providing ultra-low latency inference for open models like Kimi K3, DeepSeek V4.1, and GLM 5.3 Flash (stay tuned!). Now, we're giving agents runtime and post-training environments with Git-like control.
With LithosBox, agents can:
→ spin up, branch, and snapshot thousands of environments as they explore multiple paths forward
→ decide on a winner and rebuild the winning state
→ then throw away the environments they don't need, all in milliseconds
@dimitriosSkar, @JiaZhihao, and the rest of our team put together this agent vs. agent showdown with Doom so you can see how LithosAI helps your agents run 5x faster with AI serving infrastructure co-designed for ultra-low latency across inference and execution environments.
In our demo, the agent completes its entire run 5.29x faster with LithosBox and Lithos DeepSeek V4.1 Flash inference than it does with leading sandbox and inference providers.
Try LithosBox for free for up to two weeks via our public API.
Jon retweeted
Moving to SF isn't the hard part.
Getting into the right rooms before everyone else does is the real game.
Here's the actual map of SF founder programs worth applying to 👇
Accelerators & founder schools
-> @ycombinator - $500k on acceptance, 4 batches a year
-> @speedrun (a16z) - up to $1M, priority window runs rolling
-> @pearvc - PearX, up to $2M + $1M in credits
-> @AlchemistAcc - enterprise / B2B focused, rolling
-> @500GlobalVC - flagship program, rolling intake
-> @Berkeley_SkyDeck - UC Berkeley, semester batches
-> @draper_u - Draper University, live-in bootcamp
VC-backed residencies, fellowships & embeds
-> @sequoia - Arc, pre-seed to seed
-> @southpkcommons - $1M, the "-1 to 0" fellowship
-> @AIGrant - AI-first, $250k SAFE + heavy credits
-> @ConvictionVC - Embed, $250k uncapped, AI-native
-> @AforeVC - in Residence, backs you at idea stage
-> @unusual_vc - Unusual Academy, seed-prep program
The pre-idea programs + builder houses are coming in part 2, so follow @AroraBhavyam and turn on notifications!
Jon retweeted
Portfolio News: @webflow opens agent access in @claudeai, @ChatGPT, and @cursor_ai, plus MCP access to Webflow Analytics. The company says agents can now work with a site’s design system and CMS, make changes on branches, and bring performance data into external workflows.
Read more: webflow.com/blog/what-comes-…
Jon retweeted
Portfolio News: @harnessio Announces Capabilities that Enable Security at Machine Speed.
Harness introduced AI-powered scanning and agents that prioritize vulnerabilities, prepare fixes for review, and respond to zero-days. Virtual patching helps protect production while permanent fixes are deployed.
Read more: harness.io/blog/harness-anno…
Jon retweeted
there will always be a premium for intelligence in an instant
Jon retweeted
September’s @Unusual_VC newsletter is out. This month:
• Personal AI agents take off as Instinct and @Muse gain traction
• @Anthropic releases Claude Fable 5.1 and Mythos 5.1, with upgrades across coding, agents and knowledge work
• @OpenAI launches GPT-6 Astra, its most capable model yet
• @Chalk introduces its MCP Server and Assistant for production ML
• @Resolveai brings production context to coding agents
• AI leaders call for tools to “pace the frontier”
For more on enterprise AI and the unusual corners of it, subscribe: unusual.vc/news/
Jon retweeted
This summer, Penny Mares, our VP of Sales, had the chance to work alongside a particularly special Unusual intern: her daughter, Rose Foster.
Turns out, the learning went both ways. A great reminder that the best mentorship is a two-way street.
We loved having you on the team, Rose!
Jon retweeted
Security has always been core to Unusual Ventures. We recently brought together 50+ security leaders together to discuss what’s next.
How do we govern AI agents at runtime? Do defensive workloads really need frontier models? And how can Blue Teams share intelligence at machine speed?
The hardest questions often reveal the biggest opportunities.
Jon retweeted
Things are moving fast at LithosAI. Launching our public API (now serving both Kimi K3 and DeepSeek V4.1-Flash) has gotten a great response, and we're currently hiring for 4 engineering roles:
-Inference Optimization Engineer: Turn systems research into production performance
-Machine Learning Engineer: Make the models themselves faster while preserving full model quality
-Generalist Systems Engineer: Build the distributed systems behind fast, reliable model serving.
-Developer Relations Engineer: Grow a community around high performance model inference.
If these roles look like a fit or you know someone who’s a great fit for a team solving complex AI efficiency problems, reach out to our co-founders, @dimitriosSkar and @JiaZhihao.
This week we've also been connecting with CompSci students, other AI startups, and enterprise tech execs at @CarnegieMellon Spark 2026, @aihorizonspgh, and SCS Let's Talk (Again)! V.F26.
It’s been exciting to get more confirmation that we're focused on a problem that matters. Hearing about what our customers can build once they have ultra-low latency inference, without sacrificing model quality, and helping them see those possibilities will continue to be a big priority for us as we grow.
