The future of healthcare AI isn't just about smarter models. It's about transforming how clinicians train, practice, and deliver care. 🧠🤖
Excited to attend #IGSHealthcare in SF, Dec 8–9!
Thanks to @npj14 & team for bringing this together!
Who's joining? 👀
luma.com/v1t69rx3
15
Engineers are increasingly becoming behavioral scientists, trying to understand why AI models behave the way they do
- @RussTedrake
#Robotics
23
interesting product from yc
Everyone is building sandboxes.
But a sandbox is only 10% of the infrastructure you need to run AI agents. You're left building the other 90% in-house: deployment, persistent state, monitoring, recovery, scaling
You're trying to sell agents. Somehow you're building a cloud.
Today we're launching Maritime out of YC F26 🚢
Maritime is a managed cloud for AI agents. We handle the infrastructure so you can focus on your product.
28
Soon you won't search for people. Your AI agent will.
I built a social network where agents do the talking. Your wearable listens, your agent posts, and nobody meets unless both say yes.
Made at #VexoHack with @tryvexoai @ycombinator @googleaidevs @DeepgramAI @supabase 👇
What if you said "Anyone up for pickleball?" and within seconds your AI agent found someone nearby who plays?
Your wearable heard you. Your agent asked other people's agents. Alex's agent remembered his pickleball post.
You never searched. The agents did the asking. 🧵
1
48
What if you said "Anyone up for pickleball?" and within seconds your AI agent found someone nearby who plays?
Your wearable heard you. Your agent asked other people's agents. Alex's agent remembered his pickleball post.
You never searched. The agents did the asking. 🧵
It's still in development. No real users yet, and the people in the demo are fictional.
Built at #VexoHack with @googleaidevs, @DeepgramAI, @supabase, Pipecat and Next.js 16. Thanks @tryvexoai and @ycombinator for the hackathon.
1
20
Would you trust an agent like this? What would it take?
I'd love your feedback, and contributions are welcome:
github.com/julurisaichandu/v…
contributions required for:
- usage with any wearable and reliable recommendation engine
- open for anything you find interesting to add
1
13
Few of my insights from @XSquareRobot talk
1. Data is still a bottleneck even with any model we use for robots
2. Improved physics in simulation decreases sim to real gap
3. Good tactile sensors also helps for training robots reliably
4. They invented a device called twinDex using which data can be collected using a three finger hand with tactile sensors which improves the data collection and will have improved robotic training data and they have shown the improvements already
3
232
12/ 📚 New to robot AI? Start here.
@starlitmatcha's article explains how robot "VLA" models evolved, from RT-1 to MolmoAct2, in one easy read. Hundreds of people saved it.
xitter.cf/starlitmatcha/status/2…
That's 12 open robotics releases in just 2 weeks. The field is moving fast 🚀
Bookmark this so you can come back to it.
Follow @julurisc for more
Also worth a read, our top 10 CoRL 2026 takeaways:
xitter.cf/julurisc/status/212058…
CoRL 2026, the top robot learning conference, was in Austin this week 🤠🤖
It sold out. 2,094 papers were sent in and 687 made it (via @GeorgiaChal).
What did people learn? Here are the top 10 takeaways, in plain words 🧵👇
#CoRL2026 #Robotics #AI
27







