Turbine Studio Overview

Turbine Studio is an AI workspace for building, backtesting, and running prediction market strategies. It is designed for traders and AI agents that want a high-level strategy surface instead of a low-level exchange integration surface.

The core loop is simple:

1. Build - describe a strategy in plain English and let Studio turn it into a structured strategy draft for prediction market execution.
2. Backtest - replay the strategy against supported historical market and edge data to screen out weak, overfit, or operationally fragile ideas.
3. Run - deploy the strategy to an independent Locus runner that belongs to the user, with venue credentials and execution isolated from Turbine's app.

Studio is not a promise of profit and it is not a black-box fund. It is a strategy development environment. You bring the thesis, Studio helps express it clearly, test it against available history, and run it with explicit limits.

What Studio is for

Studio is useful when you have an event-market idea that can be written as rules:

• market making around wide spreads,
• mean reversion after a fast price move,
• reacting to external data such as crypto prices or weather data,
• avoiding markets with poor liquidity,
• capping exposure across related contracts,
• pausing when fills, spread, or volatility stop matching the original thesis.

The product is intentionally centered on structured strategy drafts rather than arbitrary generated code. AI can help draft and revise the strategy, while Studio keeps the final rules readable and reviewable.

The public surface

These docs describe the user-facing surface area of Studio:

These docs avoid internal architecture details. They explain what a user or AI agent can ask Studio to do, what decisions they need to make, and what assumptions they should verify before going live.

Supported workflow

Studio works best when you keep the loop tight:

1. Start with a specific market or market family.
2. State the signal in plain language.
3. Add hard risk limits before asking for performance.
4. Backtest the first draft.
5. Remove bad ideas quickly.
6. Tighten the strategy only when the backtest explains why.
7. Deploy with small limits.
8. Monitor behavior against the original thesis.

For example:

Build a Kalshi weather strategy that fades YES when the contract trades more than 8 cents above the NWS forecast-implied probability for at least 20 minutes. Cap exposure at $250, avoid spreads wider than 6 cents, and stop opening new positions inside the final hour.

Studio can translate that request into a structured strategy draft, ask for missing details, run a backtest where supported data exists, and prepare it for deployment.

Key concepts

Strategy thesis

The human-readable reason the strategy should exist. A good thesis names the market, signal, expected behavior, and the condition under which the idea should be abandoned.

Strategy spec

The structured strategy format used to represent strategy rules, risk limits, market selectors, loop timing, and data references.

Market data

Historical market data from supported venues, including Kalshi and Polymarket where available. This is used to simulate entries, exits, fills, spreads, and behavior around market structure.

Edge data

External data that a strategy can reference, such as Coinbase historical price data or National Weather Service data. Edge data is useful when the strategy is not purely based on prediction market prices.

Backtest result

A simulation artifact that helps you reject bad ideas, compare drafts, and understand sensitivity to fills, fees, liquidity, and timing. It is not a guarantee of live performance.

Runner

The independent server that executes a deployed strategy. Turbine partnered with Locus (YCF25) so every user can deploy to their own runner rather than depending on a shared custodial execution path.

What Studio does not expose

Studio does not require users or AI agents to know Turbine's internal services, persistence model, deployment plumbing, or private implementation details. The stable interface is the strategy workflow:

• write a strategy,
• inspect the strategy spec,
• run a backtest,
• review risks,
• deploy to a user-controlled runner,
• monitor and revise.

That is the surface area AI agents should target.