Small fixes and upkeep
3 of 4 tasks doneSee small fixes and routine upkeep together, without a delivery date.
Happening now
Coming next
Luck-resistant strategy tests
0 of 1 task doneNew tests will reject results that disappear when returns are shuffled and give agents clearer reasons for each failure.
Fair speed comparison
0 of 1 task doneYou can compare faster and slower signals on the same history to see which avoids more false switches.
Trend and range test
0 of 1 task doneOne test will use different strategies for rising, falling, sideways, and uncertain markets.
Tests on unseen periods
0 of 1 task doneCosts and results from later, unseen periods will decide whether a strategy holds up.
More markets compared
0 of 1 task doneAfter Bitcoin validation, the planned set expands to Ether, Solana, and major stock, gold, bond, and small-company funds.
Already yours
Agent-run strategy testsDelivered 9 Mar 2026
A five-iteration proof showed that an agent can change a strategy, score it, keep improvements, and revert failures.
Market condition cluesDelivered 11 Mar 2026
Three research passes gathered price, volatility, trading activity, and blockchain clues to distinguish market conditions.
Strategy filters by marketDelivered 17 Mar 2026
You can test whether trend and range filters improve a strategy without using future price information.
Three market statesDelivered 21 Jun 2026
The Python model estimates bull, neutral, and bear conditions and documents its delay and risk of fitting past data too closely.
Read-only market labelsDelivered 30 Jun 2026
Three separate measures can label trend, range, or neutral conditions without gaining permission to move money.
Live market signalDelivered 1 Jul 2026
The first read-only market signal now runs live, though it usually reacts about 67 price bars late.