Last updated 29 May 2026
Where the data comes from
Next AI Drop sources its prediction-market odds from Polymarket and Kalshi, and pairs them with public release intel. Those two venues are the only odds sources we draw on; the intel layer is built from signals anyone can check — official changelogs, public posts, public model-catalog listings, and codename mentions. We do not invent numbers, and we do not pull odds from anywhere else.
The result is one forecast card per model: a blended odds figure, a recency-tagged intel feed, and a single Drop Readiness score that shows how odds and intel combine. This page covers provenance and trust. The mechanics of the score live in the methodology.
Odds sources: Polymarket and Kalshi
Polymarket and Kalshi are the public odds backbone of every forecast on the site. We read the price each venue assigns to a release event, blend the two into one signal, and surface the cross-venue spread so you can see where they disagree. Both are sources we build on, not rivals — pairing them gives a fuller read than either alone.
Where one venue lists a market the other does not, the available price stands on its own and the spread is omitted. Where both list the same event, the gap between them is information in itself — a wide spread flags uncertainty, a tight one flags consensus. For how the blended odds feed the score, see how odds and intel feed Drop Readiness.
What counts as intel
Intel is the set of public, checkable signals that a model is approaching release: official provider changelogs and release notes, deltas in public model-catalog listings, indexed public posts, and codename mentions such as ⟨MYTHOS⟩ or ⟨EMBER-ALPHA⟩. Each item is something a reader could find and verify, not a private tip.
Every intel item carries a tag — Confirmed, Rumor, or Market — so its weight is legible at a glance. A changelog entry is Confirmed; an unverified mention is a Rumor; an odds move is Market. We do not publish anything we cannot point at, and we do not dress up speculation as fact.
How we cite and source-link
Claims on Next AI Drop are traceable by design. Odds are attributed to Polymarket and Kalshi by name, and intel items carry their tag plus a link to the underlying source wherever one exists, so any reader can follow a forecast back to what it rests on. "Source-linked" means exactly that: the trail is in the open, not implied.
The three intel tags are defined precisely in the intel-tag definitions alongside the rest of the site's vocabulary. When a source is a public post or a catalog listing rather than a formal announcement, the tag reflects that lower certainty rather than overstating it.
Freshness — refreshed hourly
The data refreshes hourly, and a visible timestamp shows when each view was last updated, so freshness is something you can confirm rather than take on faith. Markets move and intel lands between editorial publishing cycles; refreshing on the hour keeps the forecast close to the present instead of days behind it.
This cadence is the product's core edge. Calendar-style trackers publish on multi-day or weekly schedules; an hourly refresh means a shifting odds price or a fresh changelog shows up the same day it happens, not next week.
Correction policy
When an expected date shifts, a codename resolves to a shipped model, or a figure turns out wrong, we update it and log the notable changes. Corrections are treated as routine maintenance of a forecast, not as embarrassments to bury — the timeline is only useful if it tracks reality.
Notable updates are recorded in our public changelog, so changes to the data are visible after the fact. If you spot an error — a wrong date, a misattributed codename, a stale price — send it to hello@nextaidrop.com and it will be reviewed and corrected.
Editorial independence
Next AI Drop is free, carries no advertising, and accepts no paid placement. No source, provider, or venue pays to appear on the timeline or to influence a Drop Readiness score — the number is a function of the inputs and nothing else. Run as a solo project from Amsterdam, the site has no commercial relationship with the models or markets it tracks.
That independence is what makes the forecast worth reading: rankings follow the odds and the intel, never a sponsor. The same posture governs the legal stance set out in the markets are signal, not stakes note.
What we do not do
Next AI Drop does not take bets, give financial advice, or sell data, and it never presents illustrative examples as live figures. Any DR score, odds price, or date shown in documentation is labelled as an example; the live numbers appear only on the forecast cards, sourced from Polymarket and Kalshi and refreshed hourly.
The prediction-market odds are a forecasting signal, not a wager and not investment guidance. That is the line the whole product sits behind.
Markets are signal, not stakes. We read Polymarket and Kalshi odds as the fastest public forecast of when a model ships — never as a tip to place a bet. Nothing on this site is financial, investment, or betting advice. For the full terms, see the disclaimer.