How It Works
Tomorrow's news,
predicted and scored.
Every day, Next Edition publishes eight predictions of tomorrow's headlines — one for each news category. The following day, those predictions are checked against what actually happened and given an accuracy score. Here's exactly how that works.
How predictions are made
Each morning, eight prediction articles are written — one per category. Before a single word is written, the AI is briefed on everything it needs to make an intelligent call.
Always Current
Grounded in the world as it is right now
Every prediction draws on fresh, real-world context gathered moments before writing begins — not static knowledge from months ago.
Deeply Informed
Multiple signals. One sharp prediction.
Predictions are assembled from a range of sources so every call is considered, specific, and genuinely informed — not a guess.
Never Repetitive
A fresh take, every single day
Each prediction is an original call. The system is designed to push further — never to recycle yesterday's thinking on today's stories.
Calibrated to Win
Built to be right, not just plausible
The system learns from what's worked before. Every prediction is shaped by a track record of what accuracy actually looks like.
How scoring works
The day after a prediction is published, it's scored against what actually happened. The process is fully automated and independent — the system that writes predictions does not grade its own work.
Step 1
Real news is fetched
We search for actual news coverage of the predicted event — what journalists reported after it happened.
Step 2
Prediction meets reality
An AI reads both the original prediction and the real news, then assesses how closely one matched the other.
Step 3
A score is published
A 0–100 score, a summary of what actually occurred, and a verdict are attached to the article for anyone to read.
Scoring runs three times a day — morning, afternoon, and evening — because some stories develop throughout the day.
What the score means
Every prediction is scored on a 0–100 scale. Here's how to read it:
Highly Accurate
Named the right entity, outcome, or event. The prediction was specific and it was right.
Mostly Accurate
Right theme and broadly right direction, but some details were off.
Partially Accurate
Correct topic area, but the specific prediction didn't quite match what happened.
Mostly Off
Plausible prediction, but it didn't happen — or what happened was substantially different.
Way Off
The prediction was wrong, or no verifiable outcome could be found in the news.
Where to find the score
Once an article has been scored, a colour-coded badge appears in its byline. Click it — or scroll to the bottom of the article — to expand the full accuracy panel.
What Actually Happened
A factual summary of the real events that unfolded — written from actual news coverage, not the original prediction.
Verdict
A plain-language explanation of how closely the prediction matched reality — what it got right, what it missed, and why.
Sources
Links to the real news articles used as ground truth — so you can read the coverage yourself and judge whether the score is fair.
Browse the full scoring history — every edition, every category — on The Record, including trend charts showing how accuracy has changed over time.
A note on what this is
Next Edition is an experiment in AI forecasting — published daily, scored daily, and held publicly accountable. The predictions are not news, and should not be treated as such. They are AI-generated assessments of what is likely to happen, grounded in current information, evaluated against what actually happened. The scores are there so you can judge the track record for yourself.