Présage: Predict & Calibrate

Most people think they're right 90% of the time. They're actually right about 70%. The difference is overconfidence — and you can't fix it without measuring it.

Most people think they're right 90% of the time. They're actually right about 70%. The difference is overconfidence — and you can't fix it without measuring it.

Présage is a personal calibrated forecasting tool. You predict things about your own life — what you'll get done, how a meeting will go, whether the relationship will work — with a confidence percentage and a resolution date. When the date comes, you answer honestly. Présage scores you using proper scoring rules and shows you the gap between what you said and what happened.

This isn't a habit tracker, journal, or prediction market. It's a solo, honest mirror. No streaks. No badges. No gamification. The score is the score — and that's the whole point.

THE HONESTY MECHANIC

When a prediction resolves, Présage hides your original confidence and asks you Yes / No / Ambiguous first. Only after you answer does it reveal what you said. This deliberate friction prevents the most common form of self-deception: anchoring on your past confidence and finding ways to call yourself "basically right."

WHAT YOU GET

• Brier score and calibration curve, the gold standards of forecasting research • Breakdowns by category, time horizon, and mood at prediction time • Year-in-pixels heatmap of your year of forecasts • Wall of shame: your most overconfident misses • Calibration training with pre-resolved questions for fast feedback • On-device LLM coach (iOS 18+) for criteria sharpening and weekly retrospectives • Cognitive bias encyclopedia — named biases Présage watches for in your patterns • Life Forecast: 5-year vision in 5 domains, each milestone resolvable • Recurring templates: auto-spawn predictions on a schedule • Annual calibration report — printable PDF of your year, scored • Question packs by theme: career, health, relationships, world events • AI scoring: pose the same question to GPT or Claude alongside your own forecast • Twin predictions: lock in a forecast with a friend, hidden until both resolve • CSV import (Fatebook, PredictionBook) and Markdown / Obsidian export • Home Screen and Lock Screen widgets, Live Activities, watch app • Siri shortcuts for quick predict, resolve, and Brier check

PRIVATE BY DEFAULT

Présage collects nothing. Every prediction lives in a SwiftData store on your device. No analytics SDKs. No telemetry. No crash reporting. The privacy manifest declares zero tracking and zero collected data types.

iCloud sync is optional and off by default — when you turn it on, your data syncs through your private CloudKit database, never through us. HealthKit mood pre-fill is optional. Bring-your-own-key AI features call OpenAI / Anthropic directly from your device; we never proxy or log a single token.

THE MATH

Brier score, log score, and calibration bucketing follow the standard formulations from forecasting research — Tetlock's Good Judgment work, Roulston & Smith's reliability diagrams, and the original Brier (1950) paper. The Knowledge Center inside the app explains every score, every privacy decision, every gesture, and every feature in plain language.

WHO IT'S FOR

Engineers, writers, founders, scientists, traders, doctors — anyone whose work depends on judgment under uncertainty. And anyone curious enough to find out whether their gut is as good as they think it is.

The score is the score. Whether it humbles you or vindicates you — that's the whole point.