Building AI That Builds Itself

TL;DR: Today we’re releasing a new episode of our podcast AI & I. I go in depth with Yohei Nakajima, a general partner at Untapped Capital and creator of BabyAGI, the first open-source autonomous AI agent. We get into how BabyAGI has evolved, the philosophy that underpins all his creations, and his thoughts on the future of AI tooling. Watch on X or YouTube, or listen on Spotify or Apple Podcasts.

By day, he’s a general partner of a small venture firm, Untapped Capital. 

By night, he’s one of the most prolific internet tinkerers in AI. (He also sometimes works on automating his job as a venture capitalistVC.)

One of Yohei’s first viral creations was BabyAGI, an open-source autonomous agent that he built in March 2023. BabyAGI has the ability to generate and address tasks sequentially without human intervention. He’s since released seven iterations, as well as a coding agent called Ditto, a framework for building autonomous agents, and, most recently, BabyAGI 2o, a self-building autonomous agent that follows OpenAI’s unfortunate naming convention. 

I sat down with Yohei to talk about what feeds his drive to build new tools, how BabyAGI developed into a more powerful agent, and his personal philosophy around building in AI— namely, how the tools we build are extensions of ourselves. We also experiment with Ditto live on the show, building a game of Snake, and an app to help me track the names and numbers of people I’ve been meaning to schedule meetings with. And we screenshare a demo of his latest creation, BabyAGI 2o. Here’s a link to the transcript of this episode.  

This is a must-watch for anyone curious about autonomous agents, building cool AI tools on the internet, and the future of AI tooling.

Watch on X or YouTube, or listen on Spotify or Apple Podcasts. 

If you want a quick summary, here’s a taste for paying subscribers:

What drives Yohei to be a perpetual tinkerer  

Yohei loves being a VC, viewing his coding projects as an “incredibly fun hobby” that also helps his “day job.” He explains, “Some VCs will learn things and write blog posts about it—this is my version of that, but instead of blog posts, it’s code.” Here’s more about the motivations behind his hobby and how he finds time for it:

• Find creative ways to be more efficient. As a Zapier power user, Yohei has always been motivated to automate repetitive tasks, a practice that LLMs have boosted significantly. “I'm just really lazy and whenever I'm working on something that I don't like doing, I'm always asking myself, ‘How can I cut this out of my work?’”
• Unlock the hidden hours in your day. Yohei is a parent with a full-time job who makes time for his favorite hobby by using mobile versions of ChatGPT and Replit to “code in bursts” between chores. “If I'm picking up the kids and I get to school five minutes early, I can sit there for five minutes iterating on a project I've been building.”

The evolution of BabyAGI—and Yohei’s philosophy around building

Through BabyAGI, Yohei introduced the idea of “looping through an LLM”—in other words, “having an LLM generate a task list, parsing that code, and then tackling the tasks one-by-one…just using an LLM.” BabyAGI has just 100 lines of code, and Yohei believes that this “simplicity” is what made the AI tool so popular. In pursuit of sharing his ideas on what could make autonomous agents better, Yohei created seven iterations of the original BabyAGI, naming each one after a different animal.

Yohei’s experiments with a new idea—that of a self-building agent—led to the creation of BabyAGI 2, a system that provides a “framework,” or the  building blocks for “an autonomous agent that can build its own capabilities to improve itself.” His most recent release, BabyAGI 2o, is a tool that gives BabyAGI these self-building capabilities. 

This is the philosophy that underlies all of Yohei’s AI creations:

• Develop tools, develop yourself. Yohei thinks of the tools he’s building as an extension of himself, so when he’s working on them, he perceives it to be akin to working on himself. “If you take somebody who's been using a hammer for their whole life, and you scan their brain while they're using it, they almost treat it as [an] extension.”
• Reframe the way you think about your potential. Thinking about AI as an extension of himself has increased Yohei’s perception of how much he’s capable of doing. For example, due diligence tool Wokelo AI saves him time when he’s researching a new company or industry, and even though he’s “using an external tool,” he perceives it as an increase in what he can “personally get done.”
• Balance efficiency with focus. Yohei’s increased productivity has left him with conflicting emotions: feeling empowered because he's “getting more done,” yet overwhelmed as his “brain isn’t used to the throughput of information.” He recognizes the need to slow down since he “can't do too many things in parallel if [he] can't follow up correctly.”
• Find yourself in the tools you build. Beyond increasing his throughput, the exercise of building AI tools has been one of self-reflection for Yohei. For instance, while correcting an error the tool has run into, he examines his own method of solving a problem, in an attempt to “abstract the highest level pattern” and “feed it back” to the AI.  

The future of AI is personal and modular

Yohei has his finger on the pulse of what will be interesting in AI tomorrow, and these are his candid insights:

• Bespoke AI that trains itself. According to Yohei, the future of AI is personalized models that learn your preferences as you use them. “[W]hatever AI system you're using, your usage of it is going to fine-tune the underlying models that the AI system is using.” 
• Address needs while enabling growth. Yohei believes founders in AI should build solutions that address immediate customer problems while developing flexible components that can be recombined for future challenges. His recommended strategy emphasizes “customer focus [and] problem focus, but building it in a modular way.”

You can check out the episode on X, Spotify, Apple Podcasts, or YouTube. Links and timestamps are below:

• Watch on X
• Watch on YouTube
• Listen on Spotify (make sure to follow to help us rank!)
• Listen on Apple Podcasts

Timestamps:

1. Introduction: 00:00:59
2. BabyAGI and its evolution into a more powerful tool: 00:02:26
3. How better models are changing the way Yohei builds: 00:05:00
4. Using code building agent Ditto to build a game of Snake: 00:08:10
5. The ins and outs of how Ditto works: 00:13:24
6. How Yohei gets a lot done in little time: 00:19:21
7. Yohei’s personal philosophy around building AI tools: 00:21:50
8. How Yohei experiments with AI as a tech-forward parent: 00:33:13
9.  Demo of Yohei’s latest release, BabyAGI 2o: 00:39:29
10.  Yohei’s insights on the future of AI tooling: 00:51:24

What do you use AI for? Have you found any interesting or surprising use cases? We want to hear from you—and we might even interview you. Reply here to talk to me!

Miss an episode? Catch up on my recent conversations with star podcaster Dwarkesh Patel, LinkedIn cofounder Reid Hoffman, a16z Podcast host Steph Smith, economist Tyler Cowen, writer and entrepreneur David Perell, founder and newsletter operator Ben Tossell, and others, and learn how they use AI to think, create, and relate.

If you’re enjoying my work, here are a few things I recommend:

• Subscribe to Every
• Follow me on X
• Subscribe to Every’s YouTube channel

Dan Shipper is the cofounder and CEO of Every, where he writes the Chain of Thought column and hosts the podcast AI & I. You can follow him on X at @danshipper and on LinkedIn, and Every on X at @every and on LinkedIn.

We also build AI tools for readers like you. Automate repeat writing with Spiral. Organize files automatically with Sparkle. Write something great with Lex.