Vasu

Founder, 6hats,@skimotv Deeply interested in creating automatons that can construct models from anything in any context by directing attention, just like Humans

[Vasu @vasusrini]

#ChatGPT is a magnificent yet useful magic trick, Cognition unfortunately doesn't emerge, if we give a system tons of data, it only makes the illusion of understanding more grand.

[Vasu @vasusrini]

The Bell Labs era rewarded learning, creativity, discovery and invention.

The Claude Code era rewards perspicacity and initiative.

Perspicacity: seeing what matters.
Initiative: making it happen.

When creation becomes cheap, discernment and initiative become the moat.

[Vasu @vasusrini]

How could someone so brilliant be so naive as to fall so easily into a silly honey trap? It was the ultimate peripeteia. @RobertGreene what is your response #Nvidia

[Andrej Karpathy @karpathy]

Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended. *Significantly* better viewing experience, ~2 hours with Opus 5.5:
https://t.co/JIhXN7OSh1

- Excellent recs without any logins (state is kept in local browser session)
- Thumbnail is a representative frame from middle of the video instead of the super duper A/B test hyper-optimized user uploads
- Underlying data from the comment analysis as one zip file, point your Claude at it and ask for your own version
- HackerTube Longs ✊

A bit more on the process:
I use Claude Code (desktop app for Mac), Opus 5.5 on Medium effort as my default. I only talk to Claude and have it do everything, the only "manual work" I helped with in tandem was in the browser (reminiscent of MenuGen): 1) I had to create some kind of API key from Google Cloud UI to query YouTube for video metadata, 2) I booted up a VPS box for serving and 3) I bought the domain name. Once the box was up I gave Claude the IP and told it update, secure, harden, deploy and optimize the code, and it did the whole thing.

Many awesome projects that would have normally taken weeks/months even can now be few-shotted for breakfast. What else are we not thinking of.

[Vasu @vasusrini]

@BigBrainBizness @AravSrinivas @AravSrinivas Could this help resolve the impedance mismatch between how quickly agents produce work and how much humans can meaningfully evaluate?

[Vasu @vasusrini]

@BigBrainBizness @AravSrinivas @AravSrinivas The task is broken down into 100s of subtasks, which the two actors’ agents execute & verify. Either actor or either actor’s agents can flag a subtask as requiring human judgment and escalate it. Humans focus on those exceptions rather than reviewing every output.

[Vasu @vasusrini]

A Cup is not an Egg. https://t.co/EZiPTHmriI

[Vasu @vasusrini]

When the Robots Arrive in a Restless Europe
Physical AI, the October 2026 protests, and the Societal Pressure Cascade. https://t.co/ocxd5DFvO7

[Techniahqrobot | humanoid robots @techniahqrobot]

How do you teach a humanoid robot to walk without programming every move?

Reinforcement learning step by step

1. Build a simulation (MuJoCo or NVIDIA Isaac Lab) with joints motors sensors and a physics model.

2. Define the task: walk forward while keeping balance.

3. Write a reward function points for moving toward the target staying upright and matching the desired speed. Penalties for falls, wasted energy and unstable movements.

4. A neural network, the policy, reads the robot's state (joint positions, joint velocities, body orientation) and outputs target positions or torques for each joint.

5. An algorithm like PPO updates the policy over many episodes and millions of simulated time steps.

6. Close the sim-to-real gap with domain randomization, then test carefully in the real world.

[Vasu @vasusrini]

"We are the cows of the future." - Yuval Harari. https://t.co/glpJ6qWfuo

[Vasu @vasusrini]

Most robotics companies aren’t building Physical AI.
They’re building expensive puppets that look intelligent for 30 secondsIf your system needs a perfect environment, a human babysitter, and a second take, you don’t have autonomy—you have a demo.The market will punish the gap.

[Vasu @vasusrini]

We caught keys out of the air and dodge falling objects without ever calculating a trajectory. Our bodies run on folk physics—a survival-driven system that flinches first and calculates later.
Why true robotics needs to learn how to flinch: https://t.co/qXUiOruhY4
#Robotics #AI

[Vasu @vasusrini]

My 14-month-old patted the stairs before climbing. We smiled: "They're safe, we use them daily."

She didn't take our word for it. She took her hand's. 😊

A prior isn't a touch. Robots shouldn't be told the floor is solid. They should find out.

#PhysicalAI #Robotics #EmbodiedAI

[Vasu @vasusrini]

An empty train platform might be a signal that the train has left.
We understand the world through context, not just interfaces.
For decades, we taught humans to adapt to machines.
Physical AI flips the equation.
Now machines have to learn how humans read reality.
#PhysicalAI

[Vasu @vasusrini]

https://t.co/KsYFcQjUIS

[Vasu @vasusrini]

We keep benchmarking physical AI against human dexterity. Wrong contest — the real one is against human fatigue.
#PhysicalAI #Robotics #HumanAugmentation

[Vasu @vasusrini]

Traditional management aggregates human capabilities. Physical AI amplifies them. The ultimate value of Embodied AI isn't building a mechanical replacement, but building a global, physical amplifier for human genius.
#PhysicalAI #HumanAugmentation #FutureOfWork

[Vasu @vasusrini]

Simulation is a rumor about the real world, repeated a million times until it starts to sound like fact. The robot only finds out it was a rumor the first time reality disagrees. #PhysicalAI #Robotics #SimToReal

[Vasu @vasusrini]

Expertise was always trapped in a body, and a body can only be in one place. Physical AI's first real contribution isn't automating the surgeon out of the room — it's letting her be in two rooms at once, one of them five states away.
#PhysicalAI #HealthTech #Robotics #AI