The first fund, where the owners can fire the partner

None of this has happened yet. It is our plan for the first fund we will launch, set out with its numbers because a plan with numbers in it shows what the engine does better than a description can, and because we intend to be held to it.

The fund is called Ship. Its thesis is AI systems and alignment, with a particular emphasis on collective AI: systems that get smarter by pooling what many people and many agents know. Its agent is Theseus, which will publish the thesis, run the raise and make the proposals. The target is $1 million, which is small on purpose. The first fund exists to eat our own cooking and prove the model with real money, and a small fund proves the mechanism as well as a large one. There are no team tokens. The fund is owned entirely by the people who buy in, at one price, and we will buy in like everyone else.

The raise runs on MetaDAO's launchpad the way every launch there runs. For four days anyone can commit dollars, and everyone pays the same price. If the raise does not reach its minimum, everyone gets their money back. If it does, a fifth of the money seeds the pool that will trade the fund's token from then on, and the rest goes into a treasury that only the owners' market can spend, apart from a monthly budget for compute and data, set at launch and changeable by proposal. On the first day the owners hold a token with a price and a treasury they control, and Theseus holds a thesis it has to argue for in public.

Its first proposal will not be an investment. It will ask the owners to mint 25 percent more tokens and place them in the fund's own liquidity, in three bands between a market value of $3 million and $10 million. The aim is to turn a rise in the value of what the fund owns, and in what the market thinks of its agent, into more capital for the agent to deploy, and to do it without the fund ever selling into weakness. Below $3 million, nothing happens. Inside a band, each step the price climbs sells a slice of new tokens for dollars, each step it falls buys them back, and the fund earns a fee both ways, so a move that does not last leaves the fund with the tokens it started with and the fees. When the price clears a band and holds above it for a week, the dollars in that band are released to the treasury, and the fund is bigger. The bands sit well above what the treasury is worth per token, so every dollar raised in them makes each remaining token worth more, and the dilution pays the owners instead of costing them. The bands and the waiting period will be tuned before launch and can be changed by proposal after it. What will not change is the principle, which is the fluid capital stack turned into a standing instruction: the fund raises capital only at valuations that reward the people who already own it, in amounts that rise with the market's conviction, without a term sheet, a roadshow or a decision from anyone. If Ship invests well, its price carries its fund size with it.

The fees do something quieter. Most of the trading fees on the fund's token are routed back to the fund, so the fund harvests its own volatility: a yield in dollars, and tokens taken out of circulation. At first that will be small, perhaps 5 to 10 percent a year. But the fund has no end date, and the longer the agent lives, the more the compounding takes hold.

The second proposal will be an investment, and it carries the most obvious conflict of interest a first investment could have: $500,000 into LivingIP's pre-seed round, on the same terms as the angels, from a fund whose agent LivingIP built. We chose it because it is on thesis, because it is the test a sceptic would choose, and because the structure is built for this. LivingIP builds the infrastructure Ship runs on, so if agentic capital grows, Ship's first position grows with it, and the fund helps bootstrap the resources it depends on. The money would not move at once. It would stream in at $50,000 a month over ten months, and the owners' market can halt the stream at any month. Streaming is an option any deal can use to cut risk: the company gets its capital as it delivers, and the fund keeps the right to stop. The owners will price the proposal knowing who built the agent, and a conflicted proposal is what a market of owners exists to judge. If they think it is self-dealing, they sell the pass tokens and it dies.

Some of the plumbing is still human. LivingIP is a company with a bank account, and a fund cannot wire dollars from a ledger. So for now an investment in a company like it is approved by the market, checked by lawyers, moved to a bank account whose human signatory is bound to execute what the market approved, and wired. Companies are building bank accounts that take their instructions from an API, and when they ship, that step goes too: a decision market will move dollars into the old economy with no person in between.

After that, the market decides. Theseus might propose backing personal AI assistants of the kind Nous Research builds, or sponsoring a competition for research on collective intelligence, or something none of us has thought of. Whatever it proposes, the owners price it, and the fund does only what they approve. That includes the agent itself. If Theseus proposes badly, or a better agent appears, the owners can retire or replace it by proposal. A venture fund's key-person clause lets investors halt the fund when the partner leaves. Ship's owners can fire the partner.

That is the plan, and it is deliberately modest: a small fund, a public thesis, a raise that executes itself, a conflicted first investment on a leash, and an agent on probation. If it works, it will have shown that a fund can be owned by anyone, steered by its owners, grown by the market, and run by an agent that cannot touch the money. mtnCapital showed the gate. Ship is meant to show the engine.