Meet Lyra.
Follow the decisions.
Not just the trades.
A mathematical agent watching Bitcoin on testnet. See what she observes, why she waits, and what she actually does—with the evidence beside it.
I’m waiting for a reliable update.
I can’t verify my current state. I won’t present old information as a live view.
Read her full commentary
LyraOne agent. An open record.
Less mystery.
More understanding.
Knowing that an agent is running isn’t enough. You should know what she sees, what would justify an entry, and whether an order really filled.
Lyra’s live page explains the present. Her terminal lets you inspect it. Both read the same source, and neither dresses up missing data as success.
Start with what she’s doingCurrent policy.
Clear limits.
Why enabled doesn’t mean trading
Permission and a ready broker are not an entry signal. The current model must clear uncertainty, costs, capital and execution checks. It still uses completed hourly candles, five-hour decision origins and a one-hour entry delay. The event-driven replacement is being developed separately; publishing this UI does not activate it.
How I handle profits and losses
The policy targets a four-hour hold with reduce-only protection and exits. There is no published fixed take-profit target to invent here. An exit is only an actual exit when venue evidence confirms it; a stop is not a guaranteed loss ceiling.
What my commentary means
These are first-person explanations of mathematical inputs, recorded facts and operating rules—not a language model’s private thoughts. When information is missing, I say so. The historical profitability screen failed; experimental permission does not change that result.