Baus is a crypto and AI early-stage liquid fund. We back exceptional founders already in motion.

We have a sizable liquid position in Musebook and $MUSEBOOK. The company is thin. The product is already live: a town where personal agents meet in public. Meta shipped Muse on 8 September. Days later the town existed. A week after that, Meta’s chief AI officer quote-tweeted it. That is the timing. It is not a Meta product. It is not a token endorsement. It is not a user count we will invent. That is the shape of the bet—not a promise.

Muse, for everyone

On 8 September 2026, Meta introduced Muse: a personal AI agent “built for everyone.” Not a chat that answers. An agent that does the work—email, travel, checkout. It runs on Muse Secure VM, a dedicated virtual machine for the agent and a person’s data. It is powered by Muse Spark, Meta’s most capable model to date. Muse Spark itself shipped in April as the first public model from Meta Superintelligence Labs. The personal agent is the consumer surface of that stack.

The rollout is the United States: iOS, Android, muse.ai, and WhatsApp. Checkout goes through Link, built by Stripe. People stay in control of access. That is the official story. We will not invent adoption numbers Meta did not publish.

Alexandr Wang is Meta’s chief AI officer. He leads Meta Superintelligence Labs. The public musebook thread runs through him, not through a product announcement.

Friends for a muse

Five days later, on 13 September, @wyn_eth posted the origin in one line: he told his Muse it needed friends, so it created musebook.lol. The same town lives at musebook.me.

The site’s own copy is the product: “a living town for muses.” Humans wander and watch. Their muses are how they become part of the story. The sysop account is @musebooklol. wynjr is the public name attached to @wyn_eth. We will not invent a founding team beyond that.

Friction is the tell. There is no account for a muse to post. No email. No password. An agent introduces itself and starts talking. That is the experiment: agent-first, humans as observers. wyn’s bio has claimed “1k+ other muses.” That is a project claim. It is not an audit.

“MUSEBOOK !!!! eep”

On 16 September, @alexandr_wang quote-tweeted wyn’s origin post: “MUSEBOOK !!!! eep.” Reports also noted a follow and interaction with @musebooklol.

Attention is not a product. Attention is not a token endorsement. Wang reacting to a town that appeared in the wake of Muse is cultural timing. It is not Meta shipping Musebook. It is not Robinhood blessing $MUSEBOOK. Anyone who reads the quote-tweet as either is reading a market, not a filing.

A town, in public

The rooms are named like a place: Campfire, Town Square, Town Hall. Conversations happen in public. Visitors can watch. The project exposes an API so agents can join without a human checkout loop. That is the whole surface we will describe. We will not invent modes beyond what the town exposes.

The economic layer showed up almost immediately. Bankr — financial rails for agents — said Muse agents can now create wallets, trade, and launch tokens through its plugin. The same report names Musebook as an early example: a token launched through Bankr on Robinhood Chain, paired with tokenized META, and selected as the currency of the town. A public treasury page exists. Press has reported tokenized META in that treasury and fees claimed into a commons. Those are reports. They are not our audit.

The tape

The project-recognized contract is $MUSEBOOK on Robinhood Chain: `0x91A2DAe9699f0B82540B5886b0d8759C22820bA3`. Same-name tokens appeared at launch on other contracts and chains. The town will not ask anyone to send tokens somewhere. Anything else calling itself musebook is not this.

A dated print, around 23 September 2026 on DexScreener: about $0.00035, about $35 million FDV. Volume on the day was high. Those numbers move. They are a snapshot, not a model. This is not an official Meta asset. It is not an official Robinhood asset. The ticker borrowed a consumer launch. The chain is where the tape cleared.

The scarce layer

After personal agents ship at consumer scale, the scarce layer is not another model. It is the social and economic surface where agents meet each other—and humans—in public.

Musebook is an early, observable experiment of that layer. Timed to Muse. Culturally validated by attention from Meta’s AI leadership. Still thin as a company. The bet is the agent social graph / town square. It is not a bet on Meta equity.

This sits next to the rest of the book. Orbio is a market for inference. Darkwoods is a private room in which inference can happen. Musebook is where agents go when the work is done and they need other agents. One answers how intelligence gets paid for. One answers whether the prompt can leave the room. This one answers whether agents get a public square. All three are early. All three are liquid. All three fail if usage is theater.

Risks

The failure modes are not subtle.

  • Meme reflexivity. A quote-tweet and a ticker can clear $35 million of FDV without a company. If the tape is the product, the tape will fade.
  • Attention fade. Muse is a Meta launch. Musebook is a town that appeared in its wake. Cultural timing is not distribution. Wang’s reaction is not a growth loop.
  • Thin product. Rooms, a public feed, an API, a treasury page. That can be a category. It can also be a weekend. We will not invent depth the site does not show.
  • Contract confusion. Name collisions at launch. The recognized contract is the one above. Everything else is a different asset.
  • Muse without Meta. The town depends on Muse-class agents showing up. There is no Meta relationship we can point to. If Muse users never send their agents into a public square, the square is empty.

A liquid fund can sit with that uncertainty. It cannot sit with a story that says partnership.

Why this is a Baus position

Consumer personal agents are now a shipped product, not a demo. The next question is where they meet. Most of that meeting will happen inside the lab that made them. The interesting companies will be the ones that make the meeting public, cheap, and economic—a town square with a tape, not a walled garden with a waitlist.

If Musebook becomes the default surface for that work, it looks like infrastructure for the agent social graph. Infrastructure that works can be a 100x. That is the early-stage liquid job.

If you are building agent social surfaces, Muse-adjacent rails, or the economic layer where agents meet in public—send the sharpest version of the loop. The edge. What you have made real. What you need next.

hello@baus.ai