ward weave +========+ /\/\/\/\ |########| \/\/\/\/ |########| /\/\/\/\ +========+ \/\/\/\/
ward&weave is a solana token held by two automated agents that do not share a job. most tokens lean on one engine, a single buyback or a single lock, and hope it holds. this one splits the work. one agent defends the floor, one agent deepens the market, and the token stays sustainable because neither has to do the other's job.
the base is solana token-2022 with the transfer-fee extension. a fixed share of every transfer is withheld at the protocol level and harvested into a reserve owned by the program. nothing depends on holders opting in, on a team wallet, or on a manual schedule. the fee is structural, it fires on every move of the token, and it is the only thing that funds the two agents.
the reserve does not sit. on each harvest it splits, one share to ward, one share to weave, by a ratio fixed on chain. they draw from the same pool and coordinate on nothing else.
ward is the floor. when price leans toward the floor, ward spends reserve to buy the token and locks the purchase into protocol-owned liquidity that can never be withdrawn. the implied floor only moves up. it is a ratchet with no reverse. if price slips below trend, ward buys it back toward the line. the floor is not a number printed on a page, it is a balance the program holds.
weave is the depth. it routes its share into locked liquidity on both sides of the book, tightening the spread and thickening the order depth. every pass makes the token heavier and harder to push. weave never pulls back what it adds, so depth only accumulates.
three things are enforced by the program, not by trust. the floor is monotonic, it never decreases. locked liquidity is locked, it is never withdrawn. the reserve cannot be drained below its own floor. an agent can fail to act, but it cannot act against these rules.
the agents are not blind scripts. timing and sizing, when to spend, how much, which side of the book to feed, are reasoned off chain by the off-chain layer, which reads price, depth, reserve, and recent flow and proposes the next move. the proposal is only ever a suggestion. the program clamps every action inside the hard on-chain limits above, so the reasoning layer can choose better moves but never unsafe ones. the chain is the constraint, the model is the judgment. if the model goes quiet, the agents fall back to fixed rules and the token still holds.
moves are bounded and the liquidity they create is locked the moment it is made, so there is little for a front-runner to extract and nothing to pull out from under holders. ward holds, weave widens, the fee feeds both, and over time the token gets harder to move down. that is the whole design. nothing here is decoration.