A question worth forecasting.
The machine first drafts a precise, resolvable question with named outcomes. A vague worry becomes something that can be checked.
PROJECTS04
The ForecastingMachine.
Evidence-backed forecasts, from a question to a probability.
Senior software engineer / frontend leadPROJECT CONTEXT
- COMPANY
- FORAXION
- LOCATION
- London / Tunis
- PERIOD
- 2025Present
- ROLE
- Senior software engineer / frontend lead
- WEBSITE
- theforecastingmachine.com
01 / THE STARTING POINT
The future is uncertain,not unmeasurable.
The Forecasting Machine is built by FORAXION, a London company founded in 2024 as an AI-first spinoff of Hypermind, which has run prediction markets for businesses since 2000. It turns open questions into reasoned probabilities, then keeps them current against millions of articles.
Since 2025 I have led its frontend architecture: a clean white interface driven by the reading experience, developed once and shipped across web, desktop and mobile.
02 / THE PRODUCT IN PRACTICE
A probability you can read, and defend.
Evidence before the answer.
It reads the news, sorts sources into six levels of credibility and reasons with models from Anthropic, OpenAI and Google. A forecast takes four to eight minutes.
A number a leader can use.
The result travels as a PDF brief, an embeddable live widget or a public report, or through the API and alerts.
03 / A CLOSER READING
Develop once,ship everywhere.
One frontend serves the browser, the desktop and the phone, so a forecast reads the same wherever a decision is made.
Among its users: the OECD Strategic Foresight Unit, Institut Montaigne and CMAP at École Polytechnique.
One codebase
01Web02Desktop03Mobile
04 / THE THROUGH LINE
Measure the future,then make it legible.
A forecast is only useful if someone can read it, question it and act on it. The interface keeps the evidence one step away from the number.