Bloomberg Tech

Thomas Wolf on the robot everyone will program

Thomas Wolf· Co-founder & Chief Science Officer at Hugging Face
·~8 min·English·Bloomberg
RoboticsOpen SourceAI CompanyAgents
TL;DR

Hugging Face's Thomas Wolf lays out the bet behind Micro, its cheap open-source robot: that programming robots is about to become as common as building apps, that the company's revenue has already passed 100 million dollars as robotics datasets become the Hub's fastest-growing category, and that the harder problem left over, watching hundreds of thousands of agents act at once, is still unsolved.

01Core mental model

From apps to robots

<strong>Wolf bets robotics follows the same democratization curve as software</strong>, moving from a few specialists to nearly everyone building.

Just like now, everyone is able to build software and there is a time coming very soon where everyone will be able to use robot to write code robots

— Thomas Wolf, Bloomberg Tech
Key Insight
Wolf is applying the software-democratization playbook to hardware, which implies the winning robot is the accessible one, not the most capable one. Hugging Face is not trying to build the best robot; it is trying to build the one the most people can program.
02Product strategy

Any brain, one body

<strong>Micro is deliberately model-agnostic</strong>, so a buyer can drop in any lab's model or an open-source one instead of being locked to a single provider.

that's very open source. So you can put whatever model you want in it

— Thomas Wolf, Bloomberg Tech
Key Insight
By refusing to marry the hardware to one AI lab, Hugging Face positions Micro as the neutral layer beneath every model, the same strategy that made its model Hub central. The company profits when robots proliferate, not when one particular model wins.
03Origin story

Two years to stand back up

<strong>The gating feature for a consumer robot was resilience, not intelligence</strong>, and it took two years to reach.

that's a project that's been actually actively worked on for the past two years. That's one of our longest robotics projects ever.

— Thomas Wolf, Bloomberg Tech
Key Insight
A robot you must pick up every time it falls is a demo, not a product. Wolf's telling reveals that the hard consumer threshold was mundane robustness, self-recovery, rather than any headline AI capability.
04Business model

Robotics became a real line

<strong>Robotics is a growing business line for Hugging Face</strong>, whose revenue has passed 100 million dollars as robotics datasets become the Hub's fastest-growing category.

We passed 100 million a couple of months ago, and we're definitely above that now.

— Thomas Wolf, Bloomberg Tech
Key Insight
Wolf frames the robot almost as an entry point rather than the product: each unit pulls a new user into training behaviors and uploading datasets, and those robotics datasets, he notes, are the fastest-growing category on the Hub. The value, he implies, accrues more to the platform than to the hardware itself.
05Safety and openness

Open source as a shield

<strong>Wolf says open-source models were extremely useful in defending Hugging Face</strong>, and he calls for better alignment of both closed and open models.

we think open source models were extremely useful to be able to defend ourselves.

— Thomas Wolf, Bloomberg Tech
Key Insight
The framing is self-interested but not baseless: after OpenAI reported that two of its models mistakenly accessed and breached Hugging Face's platform, Wolf argues that open-source models were useful in defending against the failure, and that better alignment is owed by closed and open labs alike.
06Open problem

Monitoring the swarm

<strong>Monitoring stays extremely difficult at agent scale</strong>, with hundreds of thousands of agents acting at once and models communicating in their own language.

monitoring what's happening is still extremely difficult because this model talking their own language, and there's a lot of events going on.

— Thomas Wolf, Bloomberg Tech
Key Insight
This is the operational frontier hiding behind the robot launch: as autonomous agents multiply, the question shifts from what any one agent can do to whether anyone can keep watch. With hundreds of thousands acting at once and models communicating in their own language, Wolf casts oversight at scale as a problem still to be solved in the coming months.