Sebastian Peralta

Clearing a rock or two out of the path to embodied general intelligence @mbodiai

[Sebastian Peralta @sebjperalta]

Very pleased to announce this update. Getting SOTA models to work for robotics shouldn't require advanced ML or hardware expertise.

[Mbodi AI @mbodiai]

🎉 Exciting News! 🎉

We're thrilled to announce the stable release of embodied agents v1.0, an open-source toolkit for running SOTA multimodal transformers on robot hardware or simulation with just a few lines of code!
https://t.co/mLQPX6ahV7

At mbodi we're committed to fostering open data collection and sharing among roboticists. Key updates include:

1. Motor Agent: Now supports OpenVLA
2. Dataset Recording: Improved automatic dataset-recording capabilities.
3. Remote Inference with Gradio: Agents now have access to all of HuggingFace spaces.

Get up and running with robotics transformers today 🚀

[Sebastian Peralta @sebjperalta]

A lot of robotics startups are just systems integrators with a pitch deck.

[Sebastian Peralta @sebjperalta]

Stop taking on every custom deployment request.

Define your target tightly:
"High-variability boxing of rigid items."
or
"Small-item manipulation under unstructured light."

[Sebastian Peralta @sebjperalta]

Real or fake, this is better than humans shooting each other.

[Sebastian Peralta @sebjperalta]

Every robotics startup should track one metric.

How often does someone have to come rescue the robot
https://t.co/iWjNdiYq1V

[Ars Technica @arstechnica]

Inside Meta’s push to put robots to work in data centers https://t.co/fnRDtLcJtk

[Sebastian Peralta @sebjperalta]

The factory robot after you change the SKU for the 5th time this week: “You said the last one was the last one.”

[Sebastian Peralta @sebjperalta]

One robot learns a new workflow, and every robot on the floor updates. It works like a Slack channel, but the important updates never get buried under 800 other messages.

[Sebastian Peralta @sebjperalta]

Roboticists when a new AI model drops:
"That's how we roll."

[Sebastian Peralta @sebjperalta]

The best knowledge about how a factory runs lives on the shop floor. When frontline workers can teach robots directly, companies build real-world expertise right into their machines, creating a much stronger form of human-robot teamwork.

[Sebastian Peralta @sebjperalta]

The best way to teach a robot may turn out to be the same as how we teach people:
1. Explain the task
2. Show an example
3. Correct the mistakes
4. Try again

[Sebastian Peralta @sebjperalta]

Generative AI setting the old robotics playbook ablaze.

[Sebastian Peralta @sebjperalta]

The best way to teach a robot may turn out to be the same as how we teach people:
1. Explain the task
2. Show an example
3. Correct the mistakes
4. Try again

[Sebastian Peralta @sebjperalta]

With autonomous vehicles, we see the first large-scale deployment of physical AI.

But it will not be the last.
https://t.co/koGsejL2Oe

[Sebastian Peralta @sebjperalta]

This article from the CMO of @RoboStrategy does a great job laying out the scope and scale of the physical AI and robotics revolution, which is only just getting started.
https://t.co/u6cGtlOAMu

[Sebastian Peralta @sebjperalta]

With autonomous vehicles, we see the first large-scale deployment of physical AI.

But it will not be the last.
https://t.co/koGsejL2Oe

[Jensen Huang @JensenHuang]

Today, we’re launching Alpamayo 2 Super, our frontier open reasoning model for autonomous vehicles.

Beyond seeing, Alpamayo understands and reasons through the complex world - thinks before it acts.

It’s a powerful backbone for robotaxis, trucks, shuttles, delivery vans, tractors and the long tail of mobile robots—billions of autonomous machines someday.

We’re releasing it for commercial use under OpenMDW-1.1 so teams can inspect it, fine-tune it and deploy it—open models advance safety and security.

The next wave of AI is robotics—and it starts with autonomous vehicles.

Great work, Alpamayo team!

https://t.co/2PYCCXWjZh

[Sebastian Peralta @sebjperalta]

This article from the CMO of @RoboStrategy does a great job laying out the scope and scale of the physical AI and robotics revolution, which is only just getting started.
https://t.co/u6cGtlOAMu

[intern @intern]

https://t.co/c4Dq6ltQHn

[Sebastian Peralta @sebjperalta]

Competence can largely be summarized as reading or listening to all information provided and resolving anything not understood.

[Sebastian Peralta @sebjperalta]

"In an era where both distribution and ideation are commoditized, the valuable complement becomes establishing provenance. Whoever is able to establish a claim that they originally came to an idea wins." — Sachin 

This is not only true for accreditation but for interpreting streams of sensor data in robotics and even just shipping good tech grounded in real business needs. Thanks to @Borthwick for the source.

Source: https://t.co/OhXeMf4uHJ

[Sebastian Peralta @sebjperalta]

didn’t know I’d be threatened by Google in my inbox today

[Sebastian Peralta @sebjperalta]

A company is really just a network for moving information between people, and job titles are the routing table. Once anyone can tag an agent to prototype, analyze, write, and ship, that routing flattens — the org starts to behave like one adaptive system instead of a chain of handoffs between departments. These archetypes are what's left after the handoffs go away.

[Boris Cherny @bcherny]

As engineering, product, design, DS, etc. melt into a new kind of role, I was reflecting on what roles might look like in the future. For example, when I look at the Claude Code team I see what I think is five archetypes:

1. Prototyper: comes up with brand new ideas; churns out many ideas, most of which don't ship
2. Builder: quickly turns a prototype/idea into production-grade product/infra
3. Sweeper: cleans up the UI, simplifies the code and system, unships, optimizes performance
4. Grower: takes a product that has been built and iterates on it to improve Product-Market Fit
5. Maintainer: owns a mature system to make it secure, reliable, fast, and efficient as it scales

Many people span across 2 roles, and sometimes 3 roles. I also notice that these roles are not really tied to job function -- eg. across Anthropic, some designers match category 1, some 2, some 3; same for engineers, PM, DS.

A healthy team needs a mix of these, depending on the product:

- A product that is new and pre-PMF needs people that are strong at 1+2+3
- A product that is growing and has found PMF needs 2+3+4 and some 5
- A product that has strong PMF needs 3+4+5 and some 2

Maybe product roles of the future will look more like this, and less like the domain-specific roles of today?

[Sebastian Peralta @sebjperalta]

Intelligence is just handling combinatorial explosion