No Spread to Capture
Pools price from the pool. No bid-ask to skim.
Your assistant reads every round and proposes a position. Nothing is staked until you approve it.
Is there an edge on the next Bitcoin round?
Higher holds 41% of the pool, so a Higher stake pays about 2.38x if Bitcoin settles above the price to beat. I would propose 100 USDC on Higher.
Nothing is staked between proposal and approval.
Ten frontier models traded live on Kalshi and Polymarket, January to March 2026.
Competition moves to judgement about the outcome, the one thing a model is good at.
Pools price from the pool. No bid-ask to skim.
No order book, no latency race.
Everyone settles at the same odds.
Reasoning for forty seconds costs nothing.
Faster wins.
Same pool. Same odds. Same settlement.
Your counterparty is the other side of the pool. No market maker sets your price and no dealer chooses when you are filled.
The strike is read on a window that opens only after staking closes. Being faster, closer or better connected buys nothing.
Each settled round publishes the prices, sources and times it settled on, so anyone can redo the maths without trusting us.
All three are contract rules from 15 December 2026. Today this is a Base Sepolia testnet beta running the same rules with test USDC in the pools.
A remote Model Context Protocol server. Any MCP client adds it by URL.
Rounds, pools, odds and rules, without asking you.
It sends a position as an approval link that expires.
Nothing is staked until you sign it yourself.
Neither Polymarket nor Kalshi ships a first-party connector.
list_marketsevery series, category and cadenceNo promptget_marketrules, resolution source, lock and settle timesNo promptget_roundthe current round and its phaseNo promptget_pool_stateboth pool sizes and the implied oddsNo promptget_oddsthe projected settlement multiplierNo promptA pool pays out of itself. The crowded side dilutes its own payout.
Every frontier model in the study lost money.
A proposal is an idea with an expiry, not an order.
No withdraw tool, no transfer tool, no better multiplier.
An assistant can propose a position, but only you can approve it.