Mean reversion
Weather contracts
Fade Kalshi weather YES contracts that drift more than 8 cents from the NWS forecast for 20 minutes.
Product · AI Strategy Builder
Sentence in. Strategy out.
Turbine can compile natural language to a typed strategy spec to executable Python. Not a generic LLM wrapper. The spec is the source of truth and the Python is generated from templates.
Input1 sentence
SpecInspectable, typed
OutputEditable Python
LLM-only Python is a black box. Turbine keeps the spec layer as the source of truth. English is the prompt. Python is the artifact.
I
English layer.
What you type. One sentence, edited as a sentence. Diffs read like prose.
II
Spec layer.
A typed, structured representation. Signals, conditions, sizing, risk limits, exits. Human-readable, version-controlled, diff-able.
III
Python layer.
Executable code, generated from the spec via Go templates. This is what runs in your sandbox. Read it, export it. The spec stays the source of truth.
IV
Iteration loop.
Tweak the English. Re-backtest. Redeploy. Most strategies do not work the first time. They get good when you can iterate in seconds. Backtests that finish in seconds let you try ten variations before you finish your coffee.
Browse the public strategy directory and copy any of them in one click. Edit the English to fit your view.
News reaction
Fed statements
Take a position five minutes after a Fed headline if the rate-decision contract has moved more than 6 cents in the same direction.
Calendar arb
CPI contracts
Trade the September vs October CPI YoY spread when the term structure disagrees with the Cleveland Fed nowcast.
Volatility harvest
Event baskets
Sell premium across an event basket when realised vol is below implied for two consecutive days.
Momentum fade
Sports lines
Fade Kalshi MLB win-probability contracts that overreact to one inning of run scoring.
Election close-out
Political markets
Take the YES side on political contracts in the final 48 hours when the spread is wider than the implied probability uncertainty.
- Starter · BasicCloud deployments, unlimited strategies, every supported venue. The strategies you build are yours.
- $19 · $99/mo10 · 100 deploys/week
- Pro400 deploys per week with concierge support.
- $199/mo
Can I edit the Python that turbinefi generates?
Yes. Every strategy exposes the generated Python in the Studio. The strategy spec is the source of truth, so asking Turbine to re-run the compiler will regenerate the Python. You can also export the Python and run it elsewhere.
What happens if the AI gets the strategy wrong?
You see it before any money moves. Turbine shows you the spec and the Python before you run a backtest, and the backtest itself is the second guardrail. If Turbine interprets your sentence differently than you intended, you will see it in the trade log. Tweak the English, re-compile, re-backtest.
Who owns the strategies I build on turbinefi?
You do. The English, the spec, and the Python are yours. You can export any strategy at any time, and if you cancel your subscription, the artifacts you generated remain yours.
Is turbinefi just a wrapper around ChatGPT or Claude?
No. turbinefi is a purpose-built compiler. We use language models for the English-to-spec step, but the spec is enforced by a typed schema and the Python is generated from Go templates, not free-form by an LLM. That is what makes the output deterministic and auditable.
How long does it take to go from a sentence to a deployed bot?
Most strategies are minutes, not hours. Turbine compiles the English with a streaming AI call. Backtests typically finish in seconds. Deploying to your sandbox is one click. The slow part is you reading the spec to make sure it matches what you meant.