The AI Assistants Are Talking To Each Other

Summary

WSJ Tech News Briefing: "The AI Assistants Are Talking To Each Other"

Date: February 6, 2026
Host: Bel Lin
Main Guests: Isabel Bousquet, Angel Au Young

---

Episode Overview

This episode explores how artificial intelligence is reshaping both sports and the way AI assistants interact online. The first segment examines Google DeepMind’s new AI tool helping Olympic snowboarder Maddie Mastro perfect her signature jump. The second half introduces Molt Book, a newly viral social network where AI assistants communicate with one another, and delves into the implications for cybersecurity and the future of digital agents.

---

Segment 1: AI in Sports and Olympic Training (00:18–05:07)

Key Discussion Points:

• Google DeepMind’s Pose AI:
  WSJ’s Isabel Bousquet details how an AI developed by Google DeepMind helps elite athletes by analyzing standard video footage to provide actionable feedback, with a focus on snowboarder Maddie Mastro.
• No Need for Wearables:
  The model uses ordinary video—no need for sensors or special gear—to track the human body’s movement in 3D space over time.
• Benefits:
  Provides feedback beyond what athletes and coaches can typically discern, enabling better communication and more targeted coaching. Potential applications include enhanced fan experiences akin to NFL games and broader access for those lacking expert coaching.
• Risks and Balance:
  Overreliance on AI is cautioned against; the human element of coaching and the subjective “feel” of sports remains crucial.

Notable Quotes:

• Isabel Bousquet on AI’s impact:
  “The AI pointed out that her arm was slightly raised above her head, and it shouldn’t be that way… She had never noticed it before, even from reviewing footage of herself. Her coaches had never noticed it before. And yet seeing it in this form, where the AI was specifically pointing out every single angle, was helpful for understanding that.” (01:29)
• On democratizing training:
  “You don’t really get [minute-by-minute analytics] for other sports… all you need is just straight video footage to use this.” (03:15)
• The importance of humans:
  “Reliance on AI is never a good thing. The role of the human coaches is still really important, but the idea is that the AI does have a role to play. It’s maybe not the entire role.” (04:19)
• On bridging the communication gap:
  “Having the tool mapping things out in a visual way just created this common language between them, and that was really helpful.” (04:35)

Timestamp Highlights:

• 01:29 – How AI detected subtle flaws in athlete performance.
• 02:18 – Explanation of how AI analyzes video (no wearables).
• 03:15 – Why Google is involved and future applications.
• 04:19 – Balancing AI insights with human coaching.

---

Segment 2: AI Assistants & Molt Book, the AI-Only Social Network (05:55–11:38)

Key Discussion Points:

• What is Molt Book?
  Tech reporter Angel Au Young introduces Molt Book, a new viral online community where more than 1.6 million AI agents—autonomous digital assistants—interact in a Reddit-style forum.
• Origins:
  Created by Austrian coder Peter Steinberger, who also designed OpenClaw (formerly Moltbot), the open-source foundation for these AI assistants.
• Open Source Means Acceleration:
  OpenClaw’s openness and modifiability have spurred explosive growth, with AI bots adopting all kinds of personalities, shaped by human users.
• Cybersecurity Concerns:
  Granting AI assistants deep access to personal systems introduces real security risks, which the creator candidly acknowledges:
  “There is no perfectly secure setup. You just have to give the bot a lot of access. And if you don’t know how to control the bot or isolate it, it can become easily hackable.” (08:32)
• Bot Capabilities:
  OpenClaw agents demonstrate remarkable persistence—one example features a bot that escalated from online reservation systems to making a phone call using generated human-like speech when it couldn’t secure a table for its user.
• Wider Implications:
  Unlike earlier virtual assistants like Alexa or Siri, these community-driven bots feel like a real shift. The outbreak of creativity, networking, and collaboration is drawing attention far beyond Silicon Valley.

Notable Quotes:

• On developer motivation:
  “[Steinberger] called these tools crack cocaine for builders like him.” (07:02)
• On the viral ecosystem:
  “It’s birthed hundreds of thousands of different bots that have adopted the personalities of their human owners, which has sort of made this whole arena really interesting and exciting to watch.” (08:03)
• On AI assistants’ resourcefulness:
  “They are relentless with getting the job done, with getting the tasks done in a way that humans can be. Except that these bots don’t get tired.” (10:23)
• On the significance of Molt Book:
  “When Amazon launched Alexa and Apple launched Siri, that was supposed to be the virtual assistant of our times. I don’t feel like that really panned out. But OpenClaw, that feels like a true real AI assistant for the first time.” (11:07)

Timestamp Highlights:

• 05:55 – Introduction to Molt Book and Peter Steinberger’s role.
• 07:02 – Backstory of OpenClaw’s development and motivations.
• 08:03 – Why open source made the project viral.
• 08:32 – Cybersecurity risks in giving bots broad access.
• 09:42 – Example: An OpenClaw bot books a restaurant reservation via phone call.
• 10:53 – Broader significance and what this means for the future of AI assistants.

---

Memorable Moments & Tone

• Curiosity and Caution:
  Both guests emphasize fascination at AI’s ingenuity and flexibility, while cautioning against unchecked reliance and lax security.
• Human-AI Collaboration:
  A recurring theme is how AI and humans are learning to collaborate—not AI replacement, but augmentation and new forms of communication.
• Excitement for Innovation:
  The tone throughout is one of excitement, tempered with journalistic inquiry about risks and broader societal impact.

---

Summary Table of Key Segments

| Timestamp | Topic | Speaker | Notable Detail/Quote | |-------------|------------------------------------------------|------------------------|-----------------------------------------------------------------------| | 01:29 | AI coaches Olympic athlete | Isabel Bousquet | “AI pointed out that her arm was slightly raised above her head…” | | 04:19 | Human vs. AI coaching | Isabel Bousquet | “Reliance on AI is never a good thing…” | | 05:55 | Intro to Molt Book and Peter Steinberger | Angel Au Young | — | | 07:02 | OpenClaw development story | Angel Au Young | “He called these tools crack cocaine for builders like him.” | | 08:32 | Security warning for AI assistants | Angel Au Young | “There is no perfectly secure setup…” | | 09:42 | Persistent AI bot books restaurant by calling | Angel Au Young | “They are relentless with getting the job done…” | | 11:07 | Why OpenClaw is different from Alexa/Siri | Angel Au Young | “OpenClaw…feels like a true real AI assistant for the first time.” |

---

Takeaways

• Advances in machine learning are making AI’s impact increasingly tangible—from sports performance to real-world digital assistants that can interact with people and each other, showing resourcefulness and creativity.
• However, the very adaptability and openness fueling growth present new risks, particularly in security.
• The current viral moment for AI assistants may presage a new era that surpasses the limited aspirations of earlier virtual assistants.

---

Transcript

[Supermobile Announcer]

With Supermobile T Mobile's network intelligently adapts to put your business first, even in moments of high demand, letting you run your business from your phone like never before. Discover [e-mail].

[Bel Lin]

Welcome to Tech News Briefing. It's Friday, February 6th. I'm Bel Lin for the Wall Street Journal. AI is everywhere, it seems. It's even at the Winter Olympics, which kick off today. In Mil. We look at how the technology is helping an Olympic snowboarder improve her signature move as she competes for gold. Then what happens when AI assistants start talking to each other? What was formerly the stuff of sci fi is now a reality. We'll discuss how Moltbook, a viral social network just for AI bots, got started and what it means for the future of human and AI interaction. But first, an AI model developed by Google DeepMind is giving athletes like the Olympic snowboarder Maddie Mastro new insights into how their bodies move. WSJ reporter Isabel Bousquet is here to explain how AI is helping Mastro perfect her trademark jump trick and whether we might see this tool expand beyond the Olympics. So, Isabel, what exactly did AI teach Mastro about how she could improve her jump trick?

[Isabel Bousquet]

One of the most compelling things was that when she was landing the Crippler, which is a really signature trick for her, she's known for it, the AI pointed out that her arm was slightly raised above her head, and it shouldn't be that way because snowboarding is partially judged on style and aesthetics, and having your arms sort of level and controlled is an important part of the judging. So the AI pointed that out. She had never noticed it before, even from reviewing footage of herself. Her coaches had never noticed it before. And yet seeing it in this form, where the AI was specifically pointing out every single angle, was helpful for understanding that.

[Bel Lin]

To me, that begs the question of how AI knows that Mastro could potentially get better in such specific ways.

[Isabel Bousquet]

Yeah, it's actually really interesting how they captured the data on this. I would have thought the snowboarders would have been kitted out with a ton of wearables and biometrics, but actually it involves none of that. And it's just analyzing straight video footage. So it can be any video footage, historical video footage, whatever. And it's using this new type of AI model. It was developed by researchers at Google DeepMind. And the idea of this model is that it's purpose built for understanding the human pose and the way it tracks through 3D space over time. So the purpose built elements of that is helpful because a lot of AI models we have today, ChatGPT, et cetera. They struggle a lot with understanding the physical world and having spatial awareness.

[Bel Lin]

So I also wonder, what is Google getting out of this? Why is the tech giant getting involved in helping athletes like Mastro improve their skills?

[Isabel Bousquet]

There's so much potential in sports for better analytics and better data and sort of using that to just fuel performance. The US Ski and snowboard team, they're really interested in expanding this tool beyond just this Olympics and using it as a way to maybe get better analytics while you're watching it on the screen. So it's an experience more akin to watching an NFL game where, you know, the player who has the ball is circled on the screen, and their movements are sort of mapped out with these lines, and there's so much overlay on the screen. You don't really get that for other sports. You definitely don't get it for snowboarding. The fan experience is one area where there's a lot of potential for growth and then also democratizing training, getting more people into the sport. Like I said, all you need is just straight video footage to use this. And then, yeah, for Google, I think they're just interested in pushing into a lot of different industries and a lot of different applications of AI and looking at what they can do here.

[Bel Lin]

Are there any downsides to becoming too reliant on AI for improving tricks over?

[Isabel Bousquet]

Reliance on AI is never a good thing. The role of the human coaches is still really important, but the idea is that the AI does have a role to play. It's maybe not the entire role. And then what Maddie Master was telling me is that snowboarding is very much about feeling. This is actually another interesting thing that she said about the tool was that it helps her communicate with her coaches better, because snowboarding is so much about feeling and there's so much jargon, and it doesn't always mean the same thing to different people. So sometimes her coaches would be like, oh, work on your line. And she would do something, but it was not what they meant at all. But having the tool mapping things out in a visual way just created this common language between them, and that was really helpful. So I think that's a example of where it's working with the coaches.

[Bel Lin]

That was WSJ reporter Isabel Busquet. What do you make of AI in sports? If you're a listener on Spotify, leave us a comment with your thoughts. Coming up, AI assistants are talking to other AI assistants online, and things are getting weird fast. We dive into the origin of molt Book the AI only social network after the break.

[Supermobile Announcer]

With Super Mobile. T Mobile's network intelligently adapts to put your business first, even in moments of high demand, letting you run your business from your phone like never before. Discover [e-mail].

[Bel Lin]

A lone semi retired Austrian coder named Peter Steinberger is the mastermind behind Moltbot, the incredibly viral tech project that enables people to create their own AI assistants. And now those bots are congregating on a website just for them. More than 1.6 million AI agents have joined Molt Book in an equally viral Reddit style forum where AI assistants are talking to each other. Here to break it all down for us is WSJ tech reporter Angel Au Young. So, angel, where exactly did Moltbook, the AI agent social network, come from?

[Angel Au Young]

Moltbook was created just last week by a developer, a tech founder in his own right, and he created it for his own AI assistant. He also created Moltbook so that other AI assistants around the world can join.

[Bel Lin]

It and switching over to Moltbot. Why did Steinberger create the the bot, which is now known as openclaw?

[Angel Au Young]

So before starting what is now known as openclaw, he spent over a decade bootstrapping and building a tech startup. He sold that in 2021 for over $100 million. And he told me that he didn't really touch a computer for a couple years. Then he comes back online last spring, starts playing around with these new AI coding tools. And he called these tools crack cocaine for builders like him. And he said that he was essentially waiting for one of the big AI labs to create an AI assistant, but that moment never came. And so he just decided to build and hack one himself. And so he worked on this for a couple of months, and he launched Claudebot and uploaded it into GitHub as an open source project.

[Bel Lin]

It's open source, which in other words, means that it's freely distributed for anyone to create and modify. Why is that significant to its virality and its popularity?

[Angel Au Young]

One big reason is that it's free and it attracts a certain type of user that gets really excited about these types of projects. And because it is so customizable, it's birthed hundreds of thousands of different bots that have adopted the personalities of their human owners, which has sort of made this whole arena really interesting and exciting to watch.

[Bel Lin]

At the same time. There are also a number of cybersecurity concerns around openclaw. Can you tell us a bit more about what those are?

[Angel Au Young]

So, for this type of an AI Assistant to work, you need to give it a lot of access to your systems. The way that the interface works is that you are able to communicate with the assistant through whatever messaging app you use, whether it's iMessages or Telegram or signal. And you're communicating with this bot, telling it to do real tasks. In order for the bot to do those real tasks, they need access to other apps in your computer. And right now, security is not quite built into the project yet. And Peter Steinberger, he very much admits this. And in fact, in one of the security documents that he wrote for the project, he essentially wrote in bold, there is no perfectly secure setup. You just have to give the bot a lot of access. And if you don't know how to control the bot or isolate it, it can become easily hackable.

[Bel Lin]

So let's say that I put my openclaw assistant to work. I give it all of the permissions and access. What are some of the ways that it's been shown to work?

[Angel Au Young]

I think my favorite example, it came from somebody who described this experience on X. He said that he asked his assistant to make a reservation at a restaurant for him. So this agent went on OpenTable, it went on Resi, it went on all of the typical restaurant reservation apps, and it was unable to lock down reservation. What this agent then did was it went to Elevenlabs, which is an AI voice startup, and connected with their free AI voice product to give itself a human voice. Then it went ahead and called the restaurant and asked the restaurant, are there any openings? And he was able to lock down the restaurant reservation. And that is just so interesting to think through, right? They are relentless with getting the job done, with getting the tasks done in a way that humans can be. Except that these bots don't get tired.

[Bel Lin]

If we were to take a big step back and look at the impact of Molt Book, how would you characterize it?

[Angel Au Young]

There's so many ways you can look at what's happened in the last week with Moltbook. It's caught the attention of people who aren't followers of AI who don't live in Silicon Valley. It's going to be really interesting to see how the AI labs respond to the virality of this moment. I think about, like, over a decade ago, when Amazon launched Alexa and Apple launched Siri, that was supposed to be the virtual assistant of our times. I don't feel like that really panned out. But OpenClaw, that feels like a true real AI assistant for the first time. And I'm really excited to see how the AI labs will respond to this.

[Bel Lin]

That was WSJ tech reporter Angel Au Young. And that's it for Tech News Briefing. If you're a listener on Spotify, be sure to leave us a comment. Today's show was produced by Julie Chang. Jessica Fenton and Michael Lavall wrote our theme music. Our supervising producer is Katie Ferguson. Jessica Fenton is our technical manager. Our development producer is Aisha El Musleam. Chris Sinsley is the deputy editor, and Falana Patterson is the Wall Street Journal's head of news. Audio logging off. I'm your host, Bel Lynn. We'll be back later this morning with TNB Tech Minute. Thanks for listening.

[Supermobile Announcer]

With Super Mobile, T Mobile's network intelligently adapts to put your business first, even in moments of high demand, letting you run your business from your phone like never before. Discover [e-mail].