@BBCArdalan This might be the worst piece of music I have ever heard ๐คฆโโ๏ธ he could have just written a blog post or tweet to get his point across instead of that monstrosity of a song
Why has robotics lagged behind LLMs? On the first episode of Lab Notes, Amir Zohrenejad (@amirzo) talks with UC Berkeley PhD student and researcher Ren Wang (@ren_wang1) about world models, data bottlenecks, and what it takes to make robots work outside the lab. Tune in!
https://t.co/dZCWGst8Li
I really enjoyed hosting @ren_wang1 on the first episode of the Lab Notes podcast. Ren dove into what makes physical AI hard. Ren is world class at explaining complex technical problems in a way that even a mere mortal VC/computer scientist like me can understand.
On why simulations can't replace real-world data collection and testing:
"There is truly nothing that is higher entropy than the real world."
https://t.co/EBkJWp4gNA
On this debut episode of Lab Notes, Amir Zohrenejad (@amirzo) is joined by Ren Wang (@ren_wang1), a researcher and PhD student at UC Berkeley, to explore the state of physical AI and why robotics has progressed differently from large language models. They discuss data scarcity, world models, simulation, dexterity, and the challenges of building robots that can reliably operate in the real world.
https://t.co/mrez5WpsUj
On this debut episode of Lab Notes, Amir Zohrenejad (@amirzo) is joined by Ren Wang (@ren_wang1), a researcher and PhD student at UC Berkeley, to explore the state of physical AI and why robotics has progressed differently from large language models. They discuss data scarcity, world models, simulation, dexterity, and the challenges of building robots that can reliably operate in the real world.
https://t.co/mrez5WpsUj
At HumanLayer, weโre on a mission to solve the AI slop code problem.
In 2025 we open-sourced our Research, Plan, Implement framework, now deployed inside fortune 500s like Block and Uber - places where shipping slop is just not an option
And that was just the beginning.
Today, weโre opening access to HumanLayer - an Agentic IDE, collaboration platform, and building blocks for your software factory.
HumanLayer enables engineers solving hard problems in complex codebases to:
> move 2-3x faster across the entire SDLC (not just coding)
> maintain rigorous standards for system architecture and program design
Hundreds of engineers at companies of all sizes are already using HumanLayer to ship fast without sacrificing quality.
I'm excited to invite you to try humanlayer today at https://t.co/cQ648EkrnG, and I'm even more excited to see what you build.
@0xblacklight and I are deeply grateful to our team, our customers who give us so much incredible energy and feedback, our investors who have always been in our corner, and our friends and family who have supported us along this crazy journey
if you're a staff or principal engineer trying to make AI coding work at scale for your team, we'd love to hear from you
as @swyx likes to say - let's make this the year of no more slop
Many AI leaders in the US accused Chinese LLMs of subtle manipulation of the user (without proof, but it's hard to prove). But then the leading American lab documented manipulation of their users. Can't make this up.
๐90 days of Paperclip ๐
Itโs been just 3 months since we launched and we have:
400k downloads
69k Github Stars
10k DAUs
2k commits
90 contributors
WEโRE MOVING AT A PRETTY GOOD CLIP ๐๐๐
Thanks for building with us!
big news: i got into Y Combinator. solo founder, on my 4th attempt.
i fell in love with coding in 6th grade in Nukus, Uzbekistan, a city most people can't find on a map. since then it's been building, failing, learning, starting again: Amazon, founding engineer at a Korean unicorn, then cofounding an AI startup in Singapore.
10 years ago i sat in a San Francisco park, certain anything could be built here. now i'm back to build with the best.
Stay tuned for our launch
Agents are more expensive and less reliable than they have to be, because they're using SOTA inference for things they don't have to.
Inference is great for figuring out new workflows, naming the edge cases, and taking the wheel when things go astray. It's terrible for cheaply and deterministically running known ones. That's why we have code!
Rote captures what worked the first time so the model never has to do it twice. @heavybit is proud to have co-led their Pre-Seed w/ Seligman Ventures. Link in thread.
We raised $3M on one observation:
Agents learn on your dime. And forget.
Every run, your agent figures out the same APIs, makes the same calls, hits the same failures. Because nothing saved what worked.
We built Rote by @modiqoai to fix this.
Point it at any API. Rote reads the spec and builds a surface your agent can call: auth, endpoints, schemas. That's the adapter.
When your agent finishes a task, Rote saves the exact calls that worked. That's the flow.
Config in. Code out. Runs on your machine. Nothing to watch over.
Agents stop starting from scratch. Teams stop paying for the same work twice.
Install โ https://t.co/7clzPnKWwj
10 years ago I was sitting at the Palace of Fine Arts in SF, looking around and thinking: โdamn, this place makes you want to build something big.โ
Today Iโm back in the US on an O-1 visa.
Feels surreal.
Time to build.