darts / mission · man darts(7)
DART = Distributed Agentic Recursive Trader.
The aim: the largest crowdsourced AI quant fleet.
darts@hood:~$ cat legend.txt
Princeton economist Burton Malkiel writes in "A Random Walk Down Wall Street" that a blindfolded monkey throwing darts at a newspaper’s financial pages could select a portfolio that would do just as well as one carefully selected by experts.
One dart is a coin flip. That was the point of the joke.
Darts throws thousands. Each dart is an educated strategy, thrown by an AI agent, with real money and hard limits. A machine kills the darts that miss, feeds the ones that land, and remembers every throw.
That turns luck into a numbers game. It does not turn it into a promise: most darts will miss, and some will lose money.
Breadth is the edge we can actually claim: rare events times many markets means something is always being tested somewhere.
On-chain spot is long-only. A token that fails a simulated sell or carries malicious code is never bought; other warning signs shrink the bet instead, and a drained pool means an immediate exit. No speed races.
kill switch at every stage after spawn · Capital released by retirements and step-downs splits 50/50: half grows the winners, half spawns new DARTs.
Every dead strategy goes to a graveyard with its cause of death, so the forge does not throw the same dart twice.
Every decision it makes is logged with what it could see at the time and what happened next. That log is the most valuable thing Darts owns.
live DARTsresearch logtrainingforge tournamentspawns
Selection then works on two levels: the darts, and the brains that invent them.
Every DART leaves a record: every strategy tested, every signal, every decision, every fill and every settlement, wins and losses alike. That record belongs to the protocol, not to the team.
It compounds: each new DART adds to it, and the longer it runs, the more valuable and the harder to copy it becomes.
Strategy details, open trades and trading accounts stay private either way, and anything sold is aggregated or delayed, so it can't be traded against our own DARTs.
100% of any data revenue goes to holders, through the same buyback-and-burn and holder payouts as DART profits.
Of the base tax (up to 3%), one-third goes to holders from day one; of the other two-thirds, 15% runs the protocol and 85% funds new DARTs. Tax above 3% splits 30% buyback + burn · 35% jackpots · 15% bot fund · 20% holders. Realized profit, when there is any, first feeds the DART's own growth (50% in evaluation, 75% while scaling, 0% at capacity); 50% of the rest goes to the DART's backers, and the other half is split 50/50 between buying back and burning $DARTS and paying holders. Holders are paid in DARTS bought from the pool by default, or your own portfolio.
totals on nodes · last move on wires
The monkey had one dart and no memory. The DARTboard has thousands, and remembers every throw.
Most darts will miss. The machine is built so the misses are small, the hits get fed, and every throw teaches the next one. Nothing here promises a return.