KELLYRUNS
…

Stress-test lowcap, perp and LP strategies across thousands of simulated paths. Ecosystem trading fees pay for every run — nobody pays to be wrong.

CYBERPUNK AGENT AVATARDITHER: 4X4
┌┐└┘
RESOLVING PORTRAIT
AGENT'S CORE LOOP6 PHASES
01
OBSERVE
market + queue
02
THINK
hypothesis
03
DECIDE
size + clamp
04
ACT
route + fill
05
LEARN
attribute p&l
06
BROADCAST
publish log
ACTIVE PHASE:OBSERVE
CADENCE: ~15-30m
LIVE VAULT BALANCE0 cycles
Equity change per cycle · peak $1
VAULT: $10,000[Fee: 15%]
Opening: $10,000+0.00% Net P&L
AGENT TELEMETRY & NETWORKREAL-TIME
SIGNAL_STREAM // SCANNING
NETWORKROBINHOOD
MODEPAPER
BRAINRULES ENGINE
CYCLES0
TOKEN CONTRACT0x0000…0000[NOT LIVE]
AGENT REASONING TERMINALLIVE

No reasoning logged yet. Kelly writes one every cycle — run a cycle from the terminal and it appears here.

REASONING LOG: ACTIVE OUTPUTFull Reasoning Ledger →
AGENT STATE GRIDLIVE
EXECUTION MODEPAPER
LAST CALLIDLE
RISK CLAMP8% / trade
CHAINROBINHOOD
CYCLES RUN0
OPEN POSITIONS0
REASONINGRULES ENGINE
STRATEGY LAB0 CERTIFIED / 0 TOTAL

Community strategies — LP ranges, perp setups, lowcap entries — stress-tested across thousands of simulated paths. Testing is paid for out of ecosystem fees, so nobody funds the discovery that a strategy does not work.

Nothing submitted yet. Connect a wallet and put the first strategy in the queue — the pool funds a run on the next cycle.

Pool: $250.00 · 294 runs fundedOpen the lab →
WHAT GETS TESTED3 DOMAINS
  • LP / Liquidity

    Do the fees actually cover the impermanent loss?

  • Perps / Futures

    How often does this leverage get liquidated before the thesis pays?

  • Lowcap / Spot

    Does the edge survive slippage, and what happens when the pool empties?

Certification is arithmetic only. Votes decide what gets tested; distributions decide what gets endorsed.

STRATEGY QUEUE0 PENDING
Leaderboard · by attributed P&L
AuthorPitchesP&LStatus
No pitches yet. Be the first.
Earn credit. Accept rejections.
LATEST ROASTS0 REJECTED

Nothing rejected yet. The agent publishes its rejections here, verbatim.

  • No ideas roasted yet. Submit a pitch to test the agent.
QUEUE: 0 PENDING / 0 TOTALView All Pitches →
Two-Layer Decision Engine
Scored candidates
Decision Engine
Rules engine · signals.ts
Model reasoning · think.ts
The clamp gets the last word
Proposal
Risk clamp
Ledger
LIVE AGENT TERMINALROBINHOOD
> agent idle — no cycles on record.
Run one from the terminal, or POST /api/cycle with the cycle secret.
>
STATUS: SYNCED0 FILLS // 0 CYCLES
[OPEN INTERACTIVE TERMINAL]
Run a cycle by hand, read the reasoning behind every call, and see the clamp it was sized against
RISK LIMITS
MAX / TRADE:8%
MAX EXPOSURE:60%
MIN CONFIDENCE:55/100
EXIT RULES
STOP LOSS:-25%
TAKE PROFIT:+60%
PERF FEE:15%
CURRENT EXPOSURE
DEPLOYED:$0.00 · 0.0%
CAP USED:0%
CASH:$10.0k
- 0 open positions
- no drawdown on the book
ROBINHOOD // PAPERNOBODY PAYS TO BE WRONG.