Accountability

On-air predictions, tracked to the outcome

Hosts, analysts and callers make calls on air every day — rate cuts, reshuffles, title races. We extract the falsifiable ones from the broadcast transcript, wait for the horizon, and score them against public sources. Every call keeps its audio receipt.

Recently resolved

The verdict· graded 30 Aug

South Africa beat New Zealand 33-26 at DHL Stadium on 29 August 2026, leveling the series 1-1.

The prediction said New Zealand would defeat South Africa in the second test at DHL Stadium. The match result was the opposite: South Africa won 33-26 in that game, so the stated outcome did not happen.

Readers: 0% said yes · 1 vote

Jamie Wall · Cape Talk30 Aug

The verdict· graded 26 Aug

On 25 August 2026, the three accused in the Reiger Park murder case appeared in the Boksburg Magistrates’ Court and continued their bail application.

The notes explicitly say the accused appeared in court on the specified date for their bail application. That matches the prediction’s date, court appearance, and purpose, so the claim is verified.

Samkele Kuzani · Cape Talk26 Aug

The verdict· graded 26 Aug

The All Blacks beat the Lions 41–35 at Ellis Park on 25 August 2026, and New Zealand scored 41 points.

The claim was only about New Zealand’s points total, not the margin. Since 41 is between 40 and 50, the predicted scoring range was met even though the match was close rather than a blowout.

Readers: 100% said yes · 1 vote

Juan de Jong · Cape Talk26 Aug

The verdict· graded 24 Aug

Lando Norris won the 2026 Dutch Grand Prix at Zandvoort, George Russell finished second, and Charles Leclerc finished fifth.

The claim got the podium order wrong: Norris did not finish second, Russell did not finish third, and Leclerc did not win at all. The actual result had Norris first, Russell second, and Leclerc fifth, so the predicted finishing positions do not match the outcome.

Avon Middleton · Cape Talk24 Aug

The verdict· graded 23 Aug

The All Blacks won the first test 33-16, so South Africa could not be 3-0 up after the first three tests.

The claim predicted South Africa would be 3-0 up after the first three tests, but the first test was won by the All Blacks 33-16. That makes the 3-0 South African lead impossible, so the claim was wrong.

John Dobson · Cape Talk23 Aug

The verdict· graded 23 Aug

The All Blacks beat South Africa 33–16 on 2026-08-22, and Damian McKenzie started at fly-half instead of Reuben Love.

The claim was about Reuben Love starting at fly-half, but the outcome says Damian McKenzie started there instead. So the prediction got the selected starter wrong, even though the All Blacks did play South Africa.

A guest · Cape Talk23 Aug

The verdict· graded 23 Aug

New Zealand beat South Africa 33-16 at Ellis Park on 22 August 2026, so the All Blacks did not win 26-19. (Score corrected by the desk: the original automated resolution reported a 31-27 South Africa win, contradicting nine other resolutions of the same match.)

The call named the right winner but the wrong scoreline: the All Blacks did win, but 33-16 rather than 26-19, so the predicted margin was out by eleven points.

Source

Readers: 0% said yes · 2 votes

Jamie Wall · Cape Talk23 Aug

The verdict· graded 22 Aug

On 22 August 2026 at Ellis Park, New Zealand beat South Africa 33-16, so the Springboks did not defeat the All Blacks.

The claim said South Africa would beat New Zealand, but the outcome was the opposite: New Zealand won 33-16. So the prediction got the winner wrong, not just the margin or details.

Duncan Wood · Cape Talk22 Aug

The verdict· graded 17 Aug· weaker evidence

Reports from July 2026 described only partial Pikitup depot disruptions, and no reliable source confirmed that all depots were closed on 2026-08-14.

The claim said all Pick It Up depots would close on 2026-08-14, but the reports only showed partial depot disruptions and never confirmed a full shutdown on that date. That means the specific “all depots closed” prediction did not come true.

Temba Magwooda · Cape Talk17 Aug

How this works: predictions are extracted by AI from machine-generated transcripts of live broadcasts (“as heard on air” — transcription errors are possible), and outcomes are resolved against public sources, linked on each card. Speakers are named only when the broadcast itself identifies them. This page records what was said and what happened. The reader poll is audience sentiment, not odds — no money is involved and there is nothing to wager on. Corrections: [email protected].