The Physical AI Deployment Layer YC P26
Joined June 2026
- Tweets41
- Following19
- Followers10
- Likes25
Excited to have Synphony sponsoring DNAForge’s AI x Biology Hackathon.
We’re looking forward to seeing what problems biologists, AI engineers, and software builders can tackle together.
If you’ve been looking for a chance to test an idea, make sure to check this out.
Build at DNAForge’s AI × Biology Hackathon.
November 14th · Berkeley · $1,500+ in prizes
With @modal @boltz_bio @tamarindbio @talusbio @gxl_ai @benchflow_ai @SynphonyDev @GabriCorso @xdotli @Atelfo
Priority: October 25th
Apply in under 5 minutes:
dnaforge.com/hackathon
Synphony retweeted
Build at DNAForge’s AI × Biology Hackathon.
November 14th · Berkeley · $1,500+ in prizes
With @modal @boltz_bio @tamarindbio @talusbio @gxl_ai @benchflow_ai @SynphonyDev @GabriCorso @xdotli @Atelfo
Priority: October 25th
Apply in under 5 minutes:
dnaforge.com/hackathon
Synphony retweeted
Lessons from @lachygroom @physical_int
1. TAM for physical AI is labor on Earth, and maybe Mars
2. The right talent and capital have begun flowing into physical AI
3. PI approaches research with humility and "whatever will work"
4.Both scale and the right type of data matter
5. You can't swap one layer of the robot stack in isolation
6. Deployments are the real eval
"Whats the biggesting thing that might kill PI?"
"If our mission is unachievable"
Thank you @Mark_Goldberg_ @chemistry for putting this together!
Synphony retweeted
"Jose" starts his day picking strawberries at an organic farm. After picking he goes back to crush the berries that have fallen to the ground that are no longer usable. This prevents pest infestations and stops fungal diseases from spreading to the rest of the crop. #WeFeedYou
We made Business Insider's list of the 25 most promising robotics startups of 2026
Nominated by @LowercarbonCap
Full list: businessinsider.com/robotics…
ASHR: autonomous strawberry harvesting, powered by physical AI.
We deploy in real fields, harvesting real strawberries.
ASHR V3 Coming Soon.
We keep talking about data moats in software. Nobody's talking about the data moat sitting in physical businesses.
Every restaurant kitchen, factory floor, and repair shop generates something AI labs can't manufacture: real humans doing real physical work, all day, every day.
That's training data for the next generation of robots and most businesses are throwing it away.
A method for scaling robot training data:
Take a small real dataset (we used ~30 egocentric episodes). Run it through a world model (we used Cosmos Transfer back when it was pretty bad) to remix backgrounds and environments. You end up with hundreds of variations from the same real demonstrations. Scaled our dataset from 30 episodes to 300.
We used this to post-train our VLA models for outdoor deployment. Sharing it because more teams doing this = faster progress for physical AI as a field.
👇 What do you think is the hardest part of physical deployment? Let’s talk in the comments.
#Robotics #AI #DeepTech #Automation #MachineLearning #TechStartup #FutureOfWork #FoundationalModels
growing faster than a baby lmao
#physicalai #robotics #agtech #yc #failforward
@im_sean_wu
@lukeamlicke
@nsg_oneape