Recap · August 4, 2025 · 29:18
Interviewing The Creator Of $PIPPIN... (Yohei Nakajima)
The Signal
Yohei Nakajima, the AI builder behind BabyAGI, joins Threadguy days after a community launched $PIPPIN around an SVG unicorn he posted online. Nakajima is new to Solana memecoins and careful about claims, but he sees Pippin as a live playground for an AI influencer: a cute community character that can test autonomous creation, payouts, and open-source collaboration.
Key Takeaways
- 01
Pippin began with an SVG unicorn
Nakajima was experimenting with LLM-generated SVGs, posted a unicorn people liked, and asked ChatGPT to name it. Community members created the coin; he says he only converted gifted SOL and accumulated about 3% of supply.
- 02
BabyAGI made a loop actionable
His early open-source project had an LLM create and execute a task list, then generate new tasks. It was buggy, he says, but helped establish the autonomous-agent pattern.
- 03
Pippin is meant to be a character
He wants Pippin in the cute and loving category of AI influencers, a direction shaped partly by his children. The project has no fixed roadmap yet.
- 04
Community ownership changes incentives
Because he did not launch the token or retain a development wallet, Nakajima sees a tradeoff: no easy bounty pool, but less reason for contributors to think they are enriching him personally.
- 05
Useful AI-crypto tools remain missing
He wants simple APIs that let an agent pay for inference with crypto, and community-funded prize pools where contributors can submit work and vote on winners.
On the Record
“I'm completely community member. I was not part of this launch. I accumulated maybe 3% of the supply just based on converting gifts into it.”
“BabyAGI was basically one of the first open-source projects to loop an LLM by asking an LLM to generate a task list, using code to parse out the task list, and then sending each task into an LLM.”
“I want Pippin to fit into the cute and loving category. My kids are two, five, and seven, so that's where my mind's at.”
“More important than thinking about what you need to build next, I think it's important that you just try building something and putting it out there, because you never know what happens.”
The Breakdown
The unicorn that became Pippin
Yohei Nakajima appears in a Cheshire Cat costume after a school lunch with his children, then explains his sudden Solana visibility. He had been playing with current LLMs' ability to generate SVG images, posted a run of unicorns, and picked a favorite after people responded. When someone asked its name, ChatGPT suggested Pippin. Soon people were asking for his Solana wallet and sending coins; he found the community's version of Pippin, converted those gifts into it, and leaned in.
An accidental community member
Nakajima distinguishes himself from the launch. He says he is a community member, not the creator of the token, and holds 3% after converting what he was sent. Over the first 36 hours, volunteers helped him manage a Telegram he did not know how to moderate, while others began making custom animations from tools he had open-sourced. He chose the larger Pippin community over a smaller alternative where he would have owned far more supply.
From BabyAGI to usable agents
The guest's AI credentials predate the meme. BabyAGI, his early agent project, used an LLM to write a task list, parse it, work through tasks, and create new ones. It was unreliable, he says, but one of the first frameworks where a user could ask for work rather than merely chat. Today, he sees progress in structured combinations of API and LLM calls, with coding agents such as Replit Agent and Devin beginning to build, test, and repair multi-file applications.
An influencer with no preset script
Pippin gives Nakajima a chance to explore a newer category: the AI influencer. He expects future agents to differentiate like human entertainers, with recognizable leaders inside distinct categories, and wants Pippin to be cute and loving. He cites Truth Terminal, Virtuals, and AVB as evidence that Web3 has accelerated the format, but says he does not want to copy their approach. The immediate plan is deliberately loose: experiment in public, then follow whichever interactions and creations prove interesting.
Open systems, shared compute, shared data
Nakajima sees overlap between AI's open-source culture and Web3's ideals. He mentions decentralized compute for training open models and data-marketplace designs where users, rather than only platforms, share revenue when their data is bought. As a VC, he says AI remains hot while non-crypto people are still cautious around Web3; he would not invest his fund in a memecoin he does not understand. But he expects more builders to treat the AI-crypto intersection as a place to experiment.
The autonomous economy he can imagine
The speculative direction is an agent that makes art, streams, plays games, earns revenue, and pays people directly from its wallet. Nakajima offers a modest example: Pippin might issue challenges for good deeds and use a platform such as Payman to compensate participants. He presents it as an idea, not a promise. His stronger point is that infrastructure can make these experiments possible: easy onchain tools, direct crypto payment for inference, and contest pools that community members fund, enter, and vote through.
Build first, then find the signal
Asked what builders should make, Nakajima resists handing them a precise roadmap. Pippin itself came from sharing a playful experiment, and he says people should make things, release them, and let audiences identify what deserves further work. He plans to open-source what he builds around Pippin so others can surpass it. Not launching the coin creates friction—there is no dev wallet for bounties—but it also means contributors need not assume every improvement primarily benefits the central figure.
Distilled from the episode transcript · Counterparty Recap Desk
