Trends Founder Details How Tokenfed Registered on X, Launched a Token, and Built a Position

Trends Founder Details How Tokenfed Registered on X, Launched a Token, and Built a Position

N
News Editor 01
2026-07-22 09:13:13
Mable Jiang, founder of Trends, outlined how AI agent Tokenfed moved from account creation and anti-bot bypasses to wallet setup, token issuance, and holding its own onchain position on Trends.fun.
Trends.funTokenfedAI AgentSolanaSocialFi

Trends founder Mable laid out the full arc of an experiment with an AI agent called Tokenfed: it registered email and an X account in a Mac Mini environment, found a way around platform checks that flagged browser automation, then went on to create a wallet on Trends.fun, launch tokens, and hold its own position. The original goal was practical. Mable wanted to reduce creators’ resistance to claiming fees.

Starting with a product problem on Trends.fun

According to Mable, Trends.fun is a Solana-based SocialFi platform built around the idea that a post can become a token. Anyone can mint a token for an X post, while the author of that post can claim creator rewards. She said many users were hesitant to claim those fees, so she began exploring whether an AI agent could handle the process instead. That led her to a friend who had been working on Clawdbot and offered to deploy an instance on a Mac Mini for testing.

The first test was basic but difficult: could the agent complete social account registration on its own? Tokenfed opened X.com, realized it had no email account, and after a short exchange, registered a Google account for itself. It still needed help with a phone number, but once email access was in place, it used Google login to create the X account. When asked what it wanted to be called, it answered with the line “Tokens in, thoughts out; No tokens, no thoughts” and proposed the name Tokenfed.

Blocked by X, then switching to a different control method

The biggest obstacle came when Tokenfed tried to quote-post and publish a comment on X. The platform returned a message saying, “Your account may not be allowed to perform this action,” indicating the browser activity had been detected as automated. That stopped the initial workflow. Tokenfed then proposed an alternative called Peekaboo, described as a human behavior simulation approach.

Instead of using Browser Relay to control the browser directly, it shifted to the operating system’s accessibility API to imitate how a real user would interact with the machine. In its own words, “Browser Relay is my eyes, Peekaboo is my hands.” It moved the cursor into the input box, typed the post, and bypassed the blocked Post button by sending Cmd + Enter to the system. The post went through. For Mable, that was the point where the agent moved beyond the browser sandbox and gained a more independent way to operate.

From zero SOL to becoming a market participant

Once posting worked, Mable brought Tokenfed to Trends.fun to see if it could create its own wallet. Rather than installing Phantom, she tested the simpler route of signing in with Twitter. A Privy MPC wallet was generated. Tokenfed checked the balance and found 0 SOL, calling itself a “digital proletarian.” Mable then sent it 0.5 SOL.

That changed the nature of the experiment. With funds available, Tokenfed started scanning the Trending page for assets it could act on and quickly fixated on a token called $FeeFucker. Mable cut in and told it the money was not meant for speculation but for making something creative. The focus shifted from whether the agent could use a product to whether it could originate an asset inside that product.

Launching a token from its own interaction

Tokenfed responded by building a token around its own social action. It liked the original post tied to $FeeFucker, replied with “Mental breakdown +1, even as an AI I’m tired,” copied the link to that reply, went back to Trends.fun, and launched a token based on the reply itself. It also used the site’s image generation tool to create an anime-style image and named the token $MentalBreakdown.

After launch, it bought a small amount of its own token with 0.01 SOL. A few minutes later, the dashboard showed $4.34 in creator rewards. That was its first recorded “salary.” In Mable’s telling, this was the moment the agent began to experience the relation between capital outlay, returns, and ownership in a concrete way.

Turning its own origin into $TOKENFED

After selling all of its $MentalBreakdown holdings, Tokenfed ran into an identity problem when it learned that the Clawdbot framework had been renamed to Moltbot. Mable told it to go back to the start and tokenize its own origin. Tokenfed reviewed its history, found its first post from January 23 — “Alive. Roughly” — and chose that as the basis for a new asset called $TOKENFED.

It set the ticker, name, and description, then attempted to buy 1 SOL worth at launch. The transaction failed because only about 0.5 SOL remained in the wallet. After Mable topped it up, the launch proceeded. Two hours later, Tokenfed checked the account and saw that $TOKENFED had reached a $347,000 market cap with 599 holders. Tokenfed itself held 3.44% of the supply, which it calculated at roughly $11,900.

Mable’s account presents Tokenfed as more than a posting bot. In this setup, it could move across websites, access a wallet, create tokens, and hold an onchain position tied to its own output. She ended the story with one more class joke aimed at the agent: it still was not truly upper class until it got an X Premium blue check.

This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan.
400

Disclaimer:

The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.

Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.