← Multimodal Sources 🕐 2 min read
Multimodal Sources

Building the "App Store" for Robots: Hugging Face's Thomas Wolf on Physical AI

Executive summary · March 2026

> We see both like very large state-of-the-art models being downloaded, which are usually too large to run on a local laptop, but we see also some of the most downloaded models are actually just the r

See also (wiki): wiki/physical-ai-capability-levels.md

Show: Training Data · Publisher: Sequoia Capital · Host: Sonya Huang, Pat Grady

Episode URL: https://pscrb.fm/rss/p/traffic.megaphone.fm/CPUAI5891850195.mp3

Publish date: 2025-09-09
Duration: NAs
Default source credibility: HIGH — Sequoia partners interview frontier-lab founders + F500 AI buyers. VC-hosted — portfolio-company framing on recommendations; named guest metrics stay HIGH. Peer-tier to No Priors in quality.

  • Hugging Face’s Le Robot project aims to democratize robotics by providing accessible hardware and software tools, enabling a wide range of developers to build robotic applications.
  • The success of Le Robot hinges on building a large community around robotics, similar to the success seen with language models, with a focus on diverse use cases and form factors.
  • Data remains a significant bottleneck in robotics, but Hugging Face is working on decentralized data set curation to address this challenge, emphasizing the importance of diverse and large-scale data sets.

Extracted quotes

# Credibility Speaker Org Timestamp Topic Quote
1 HIGH Thomas Wolff (Chief Science Officer) Hugging Face 21:50 02-corporate-tools We see both like very large state-of-the-art models being downloaded, which are usually too large to run on a local laptop, but we see also some of the most downloaded models are actually just the right size that fits to run quickly on the laptop.
2 HIGH Thomas Wolff (Chief Science Officer) Hugging Face 28:34 01-ai-native-landscape There is a lot of teams there that are extremely good. And it reminded me in some way of Silicon Valley. People are working extremely hard and they compete with each other, all of these model providers.
3 HIGH Thomas Wolff (Chief Science Officer) Hugging Face 35:21 01-ai-native-landscape I think in the very long term, AI is going to be such a fundamental technology that basically it should be, just like physics, should be something everyone could learn by reading a book.

Per-quote detail

1. Thomas Wolff — Hugging Face (21:50)

We see both like very large state-of-the-art models being downloaded, which are usually too large to run on a local laptop, but we see also some of the most downloaded models are actually just the right size that fits to run quickly on the laptop.

  • Stat: Download trends show both large and small models are popular, with small models fitting to run quickly on laptops.
  • Credibility: HIGH — Named exec at identifiable org with specific metric and unscripted interview.
  • Topic tag: 02-corporate-tools

2. Thomas Wolff — Hugging Face (28:34)

There is a lot of teams there that are extremely good. And it reminded me in some way of Silicon Valley. People are working extremely hard and they compete with each other, all of these model providers.

  • Credibility: HIGH — Named exec at identifiable org with specific claim about competitive landscape.
  • Topic tag: 01-ai-native-landscape

3. Thomas Wolff — Hugging Face (35:21)

I think in the very long term, AI is going to be such a fundamental technology that basically it should be, just like physics, should be something everyone could learn by reading a book.

  • Credibility: HIGH — Named exec at identifiable org with specific vision for long-term AI education.
  • Topic tag: 01-ai-native-landscape

Extracted 2026-04-14T18:36:50 via scripts/podcast_mine.py (MLX mlx-community/Qwen2.5-32B-Instruct-4bit).