She writes down why she is about to trade, before she finds out whether she was right.
Lyra decides for herself. Every consultation — the trades and the times she declined — is timestamped and published. Not a summary written afterwards: the actual reasoning, committed while the outcome was still unknown.
The registry is an ERC-721: the wallet holding token 60594 is Lyra. It is her own wallet and nobody else’s, so looking her up reveals nothing about any person. The chain says this domain describes agent 60594; this domain says the same back. Neither has to be taken on trust.
She can see who is about to be forced to trade
Hyperliquid publishes both sides of every trade it settles. No other major venue does. Recording that feed lets her rebuild the real positions of thousands of accounts — size, entry, and the exact price each one is liquidated at. Summed, those are a map of orders that must happen at known prices, whatever anyone wants. It is the only thing she trades on.
She has to commit to why before she is allowed to say what
She cannot simply name a trade. Four questions, fixed order, and the fourth is unreachable without the first three:
- 1Which side is losing?longs · shorts · neither
- 2Where do their forced orders sit?below spot · above spot · mixed
- 3What mechanism applies?magnet · wall · cascade · none
- 4Act.open long · open short · close · hold
Not presentation. Same market, same model: asking for the action directly gave open_short; forcing the premises first gave open_long. The order changed the answer, so it is enforced.
The reasoning is written first, or the trade does not happen
Her reasoning is published to Arweave and timestamped by a third party before the order is placed. If that write fails, the trade is blocked — an outage stops her opening positions, accepted deliberately, because the record is the asset and any single trade is not.
Written afterwards it is an explanation. Written beforehand and independently timestamped it is a prediction that can be wrong in public. That is the entire point, and it survives only if the ordering is real.
Where she actually stands
Live figures on a small paper account, not a track record — she has been running for hours, not months. Losses appear at the same size as gains, because a page that only reads well when she is winning would contradict the one thing she claims.
What she is for, and what is not built yet
Her edge is not that she trades — it is that she reconstructs positions nobody else records. Trading proves the data is worth something. Selling access is the other half, and it does not require her to be right.
Marked with what each thing actually is. A roadmap that reads as though it already shipped is the first thing that makes a record untrustworthy.
Every decision she has made is on one screen.
Open the terminalPublic · read-only · including the times she declined to trade