Could AI have predicted the Trump–Putin diesel deal?
We sent Sikt back to October 5, four days before the Trump–Putin diesel deal. It mapped the exact path, then gave it 2%. Here is what it saw and missed.
Read the articleSikt's AI forecast for the 2026 midterms: the chance each party wins the Senate and the House, a pick in ten key Senate races, and where it differs from the markets.

Short answer: Sikt's AI forecaster favors Democrats to win both the House (90%) and the Senate (65%), in its latest forecast (round 7, from Oct 21, 2026). It also makes a call in ten key Senate races. The numbers below come from the Sikt Midterm Bench, our public test of an AI forecaster on the 2026 midterms, and they update with every new round until election day on November 3.
A live AI forecast, scored in public. In the Sikt Midterm Bench, Sikt forecasts the 2026 midterms next to Kalshi and Polymarket. Latest round, Oct 21, 2026: Senate: Democrats 65% (markets 68%), every round; House: Democrats 90% (markets 91% to 92%), every round. Every forecast will be graded after election day, misses included.
Sikt Intelligence is an AI forecasting startup from Oslo, Norway. Our AI forecaster researches a question the way an analyst would: it reads the polls, the news, the history of similar races and the base rates, then puts a probability on the outcome and writes down its reasoning.
The 2026 midterms are its first big public test. Since September 30 we have published a dated forecast every few days for control of the Senate, control of the House and ten key Senate races. Every round stays on the record, so you can see how the numbers moved and, after the election, how well they held up.
A forecast that is dated, public and graded afterwards is worth more than a confident take nobody checks.
These are Sikt's picks in the ten races that decide the Senate, closest first. "Sikt" is the AI's chance that its pick wins; "Markets" is the average price for the same candidate on Kalshi and Polymarket. Click a state for Sikt's reasoning, every round and the live odds.
| Race | Sikt's pick | Sikt | Markets |
|---|---|---|---|
| Maine | Jackson (D) | 54% | 66% |
| Iowa | Hinson (R) | 55% | 57% |
| Alaska | Peltola (D) | 57% | 74% |
| Texas | Talarico (D) | 57% | 66% |
| Ohio | Brown (D) | 58% | 70% |
| Nebraska | Ricketts (R) | 60% | 74% |
| Michigan | El-Sayed (D) | 71% | 70% |
| New Hampshire | Pappas (D) | 87% | 90% |
| Georgia | Ossoff (D) | 93% | 96% |
| North Carolina | Cooper (D) | 95% | 96% |
The seat math. Republicans hold 53 seats and Democrats 47, counting the two independents who caucus with them. Democrats need a net gain of four seats for a majority, because a 50-50 Senate stays Republican with Vice President JD Vance's tie-breaking vote. If every Sikt favorite in the 10 races won, Maine, Alaska, Texas, Ohio and North Carolina would change hands, for 52 Democratic seats and a Democratic majority, assuming the other 25 races go to the party that holds them.
That count is a simple sum, not the forecast. Races move together: a national swing lifts or sinks a party's candidates in many states at once, so the chance of winning the Senate is not the product of the race odds. That is why Sikt also forecasts Senate control directly. Sikt's latest Midterm Bench forecast, from Oct 21, 2026, gives Democrats a 65% chance of controlling the Senate, against 68% on the markets. See every round on Sikt's Senate forecast page.
Sikt and the markets differ most in Alaska (Mary Peltola: Sikt 57%, markets 74%) and Nebraska (Pete Ricketts: Sikt 60%, markets 74%). In the other 8 races the two are within 12 points. In 9 of the 10 races Sikt is less sure of the favorite than the markets are.
Disagreements are the most interesting part of the test. When the AI and the markets agree, both are probably reading the same polls. When they differ, one of them is weighing something the other is not, such as a state's voting history, a ranked-choice count or an independent candidate who does not fit the usual two-party pattern. The election will show which one read it better. For the full picture of every race, including the candidates and how close each contest is, see who will win the Senate in 2026?
Sikt's latest Midterm Bench forecast, from Oct 21, 2026, gives Democrats a 90% chance of controlling the House, against 91% to 92% on the markets. See every round on Sikt's House forecast page.
More on what decides the House in who will win the House in 2026?
Most midterm predictions you will see this fall come from one of three traditional sources. Each answers a slightly different question.
| Source | What it measures | Strength | Weakness |
|---|---|---|---|
| Polls | What voters say today | Direct evidence from voters | A snapshot, not a forecast; misses in both directions |
| Forecast models | Polls and fundamentals, simulated thousands of times | Consistent and transparent | Only as good as the polls and the assumptions they are built on |
| Prediction markets | The price traders put on each outcome | Fast to react to news; money on the line | Thin markets and long shots can be mispriced (why) |
| AI forecaster | Everything above, plus news and history, read like an analyst | Explains its reasoning; can cover any question | New, with a short public track record |
An AI forecaster does not replace polls. It reads them, together with the models, the markets, the news and the history, and then makes its own call. Whether that adds anything is exactly what the Bench is testing. The evidence so far is in can AI predict elections? and AI vs. superforecasters.
After the results are in, every round is scored with the Brier score, the standard measure for probability forecasts. It rewards being right and confident, and it punishes being confidently wrong. We will also check calibration: races Sikt gave about 60% should be won by its pick about six times in ten, not ten times in ten.
A good forecaster can still lose a race it gave 60%. That is what 60% means. The test is the whole set of forecasts, every round, misses included.
Possibly not on election night. Ranked-choice counts in Maine and Alaska can take days, and a Georgia race that goes to a runoff would be decided weeks later. In 2022, Senate control was settled on November 12, four days after the election.
Sikt's AI forecaster favors Democrats to win both the House (90%) and the Senate (65%), in its latest forecast (round 7, from Oct 21, 2026). On the markets, prediction markets favor Democrats to control the Senate (Kalshi 68%, Polymarket 68%), and prediction markets favor Democrats to control the House (Kalshi 91%, Polymarket 92%). The numbers in this article update with every new round.
Sikt's latest Midterm Bench forecast, from Oct 21, 2026, gives Democrats a 65% chance of controlling the Senate, against 68% on the markets. See every round on Sikt's Senate forecast page. Democrats need a net gain of four seats for a majority.
Sikt's latest Midterm Bench forecast, from Oct 21, 2026, gives Democrats a 90% chance of controlling the House, against 91% to 92% on the markets. See every round on Sikt's House forecast page. A House majority takes 218 of 435 seats.
That is what the Sikt Midterm Bench is testing in public. An AI forecaster reads polls, news and history and puts a probability on each race. Its forecasts are dated and will be graded against the results after November 3. See can AI predict elections? for the evidence so far.
Tuesday, November 3, 2026. All 435 House seats and 35 Senate seats are on the ballot.
Every few days. Each round is dated and kept on the record on our 2026 midterms page, so you can see how the forecast moved.