Manuel Yves Galliker

Learning, Robotics, Controls & Open Source Enthusiast

[Manuel Yves Galliker @manumerous]

Over the last few years, end-to-end robot learning has moved incredibly fast. Even working at the frontier, I’m regularly surprised by how much it can already do. And yet we still haven’t seen it deployed at scale.

At @enact_intel, we think that’s because no one has fully committed to this new paradigm yet. Most approaches fall into one of two camps:

Humanoid companies start with the form factor. The body comes first, built to look like us, and the AI gets adapted to hardware designed before the model existed.
Pure model companies run on off-the-shelf hardware. Fast and convenient, but it limits the modalities and the quality of the data.

We see it differently: the body shapes the data, and the data shapes the model.

So we’re rebuilding the whole stack: model architectures for multi-frequency, multi-scale intelligence, the data pipelines, and all the way down to the actuators and sensors feeding the model. Our robots sense force and touch through their own actuators, so they feel what’s happening in real time. That makes learning more data-efficient while pushing reliability and performance. It works across form factors and is built for real-world deployment from day one.

Huge thanks to everyone who’s helped along the way: early supporters, everyone who took time to give advice, our angels, and the fantastic team I get to tinker on robots with late into the night.
@redalpine @firstminutecap @Notion_Cap @ylecun @AndrewLonsberry @arthurmensch @Thom_Wolf @mapo1 @HaraldNieder @spencer_crawley @ascribblenau7 and many more...

[Manuel Yves Galliker @manumerous]

@_virajacharya @enact_intel Anytime!

[Manuel Yves Galliker @manumerous]

Over the last few years, end-to-end robot learning has moved incredibly fast. Even working at the frontier, I’m regularly surprised by how much it can already do. And yet we still haven’t seen it deployed at scale.

At @enact_intel, we think that’s because no one has fully committed to this new paradigm yet. Most approaches fall into one of two camps:

Humanoid companies start with the form factor. The body comes first, built to look like us, and the AI gets adapted to hardware designed before the model existed.
Pure model companies run on off-the-shelf hardware. Fast and convenient, but it limits the modalities and the quality of the data.

We see it differently: the body shapes the data, and the data shapes the model.

So we’re rebuilding the whole stack: model architectures for multi-frequency, multi-scale intelligence, the data pipelines, and all the way down to the actuators and sensors feeding the model. Our robots sense force and touch through their own actuators, so they feel what’s happening in real time. That makes learning more data-efficient while pushing reliability and performance. It works across form factors and is built for real-world deployment from day one.

Huge thanks to everyone who’s helped along the way: early supporters, everyone who took time to give advice, our angels, and the fantastic team I get to tinker on robots with late into the night.
@redalpine @firstminutecap @Notion_Cap @ylecun @AndrewLonsberry @arthurmensch @Thom_Wolf @mapo1 @HaraldNieder @spencer_crawley @ascribblenau7 and many more...

[Manuel Yves Galliker @manumerous]

@lukas_m_ziegler Thank you so much for sharing Lukas!

[Lukas Ziegler @lukas_m_ziegler]

🚨 BREAKING:

Enact has emerged from stealth with $30M in pre-seed funding!

The largest pre-seed round ever raised in Switzerland! 🇨🇭

Backed by redalpine, Notion Capital and firstminute capital.

Angel investors include Mistral founder Arthur Mensch, Yann LeCun, and Andrew Lonsberry.

The company is building robots that feel force and touch, but not through external sensors.

Through their motors. Touch sensors exist. But they tend to be too costly and fragile for real factory conditions.

Enact's approach measures force inside the motors themselves, comparing it to the way humans feel tension in their muscles, not just at their fingertips.

No separate sensors to break. No additional hardware to integrate. The capability lives in what the robot already has. And it works across anything from a simple robot arm to a full humanoid.

Founder @manumerous built the software for 1X's NEO humanoid before joining Physical Intelligence.

He founded Enact in March 2026. The 12-person team includes engineers from DeepMind, Amazon Robotics and Helsing.

He chose Zürich over Silicon Valley deliberately.

"Silicon Valley has traditionally been very strong in software and AI, but it's lacking in robotics and hardware engineering talent."

Plans to grow the team fivefold by end of 2027.

First deployments in manufacturing, infrastructure, retail and food processing across Europe.

Touch has been the missing sense in robotics for decades. Enact is building it into the motor itself. 🫆

~~ 

♻️ Join the weekly robotics newsletter, and never miss any news → https://t.co/GoA3ZuwWF9

[Manuel Yves Galliker @manumerous]

@ericjang11 Love it!

[Sourish Jasti @SourishJasti]

1/ General-purpose robotics is the rare technological frontier where the US / China started at roughly the same time and there's no clear winner yet.

To better understand the landscape, @zoeytang_1007, @intelchentwo, @vishnuman0 and I spent the last ~8 weeks creating a deep dive on humanoid robotics hardware and flew to China to see the supply chain firsthand.

Here's everything we've created + our takeaways about the components, humanoid comparisons, supply chains, and geopolitics👇

[Axel @CoRL Austin ✈️ @ax_pey]

Introducing MARS, the first Personal AI Robot.

In technology, getting to the future we want starts with building it together.

This is why we designed Mars. Inspired by early PCs, Mars is a powerful, complete, extendable robot.

It runs BASIC, an open embodied AI agent we created for you to program with code, demonstrations, and prompts. This allows Mars to perform complex long-horizon tasks involving spatial memory, reasoning, manipulation and navigation.

Pre-orders open now

⬇️ Demos & details in the thread

[Manuel Yves Galliker @manumerous]

@ericjang11 Man walking over legos hurts! 😂

[Manuel Yves Galliker @manumerous]

@sadusumi @eigensteve Hi Srinivas, would love to follow up on this and see if I can help 🙂

[Manuel Yves Galliker @manumerous]

@eigensteve is easily one of my favorite resources to learn or revisit topics from ML amd Controls. Can't wait for the new series!

[Steven Brunton @eigensteve]

New Book & Video Series!!! (late 2025)

Optimization Bootcamp: Applications in Machine Learning, Control, and Inverse Problems

Comment for a sneak peak to help proofread and I'll DM
(proof reading, typos, HW problems, all get acknowledgment in book!)

[Manuel Yves Galliker @manumerous]

@breadli428 Yeah people learn the sae lessons all over again 😁

[Minn @minney_cat]

this year alone, I've met hundreds of the world's elite AI researchers + engineers steering the future of intelligence, and they ultimately want to do it here, in the US 🇺🇸

If a visa or green card are holding you back, join us on 7/31 in SF and hear real stories from @MatternJustus, @manumerous, @theaievangelist, and me @lighthousehq_ on getting unblocked to pursue their ambitions

RSVP in thread

[Manuel Yves Galliker @manumerous]

@sadusumi @eigensteve That sounds interesting. would you like to discuss the details offline? (manuel[at]galliker{dot}tech)

[Manuel Yves Galliker @manumerous]

@FenglongS @breadli428 @zhaomingxie Nice to hear from you @FenglongS! If you formulate the problem correctly and optimize through the WB kinematics/dynamics getting a straight knee gait is doable: https://t.co/05bmxjqJPt I think for reduced order models that is more tricky yes. 😀

[Manuel Yves Galliker @manumerous]

@sadusumi @eigensteve What exactly do you have in mind?

[Manuel Yves Galliker @manumerous]

@sadusumi @eigensteve Ah yes definitely!

[Manuel Yves Galliker @manumerous]

@breadli428 Great share! I think this combination makes a lot of sense since it play into the strength of both approaches (planning for long horizon tasks for MPC and unparralelled robustness for RL)! In fact I think we would ideally combine the advantages of both approaches into one!

[Manuel Yves Galliker @manumerous]

@breadli428 @zhaomingxie I think what is really holding MPC robustness back is the assumption of a known contact sequence with a mismatch leading to catastrophic failures. Optimizing for this means solving an online mixed integer problem. RL has less assumptions there and does that computation offline 😁

[Manuel Yves Galliker @manumerous]

@breadli428 @zhaomingxie Interesting take. As someone that has extensively worked on both RL and WB-Dynamics MPC for legged locomotion I rarely had the feeling that the costs/rewards was the bottleneck for MPC. In fact they seemed remarkably similar (track base velocity, regularitations, min torques).

[Manuel Yves Galliker @manumerous]

@SpaceWarriorX @adcock_brett Likewise! I think figure's policies do look already pretty smooth as is though!