AI Market Intelligence Daily

1. The single biggest move in the entire dataset is a structural repricing on the Anthropic model-launch ladder, not a headline event. On Kalshi, "Will Anthropic release next Mythos-Class model before Oct 1, 2026?" jumped 52% to 84%, and the Nov 1 window climbed 63% to 92%, while the Sep 1 window collapsed 14% to 6%. Together these three contracts imply the market has suddenly pinned the next Mythos-Class release into the late-September/early-October window with high confidence. That is the clearest signal of the day: traders are not just expecting a release, they are concentrating it into a specific two-week window.

2. This concentration is corroborated by the adjacent Anthropic rungs, which all moved in the same direction. "Opus 6 before Nov 1" fell 16% to 9% (a Mythos-Class launch would crowd out Opus 6), "Opus 6 before Oct 16" rose 6% to 8%, and "Mythos before Dec 1" rose 13% to 16%. The fact that the Sep 1 Mythos contract dropped sharply while Oct 1 and Nov 1 surged is the tell — this is a re-dating of an expected event, not a change in the probability of an event at all. It looks like a scheduled launch was leaked or announced as landing around late September.

3. The Google Gemini cluster moved in the same re-dating pattern. "Gemini 3.5 Pro before Aug 21" dropped 7% to 2% (the near-term window is now considered nearly dead, with only days left before expiry), while "Gemini 3.5 Pro before Aug 31" rose 9% to 11%. The market is pushing the Gemini 3.5 Pro launch from the immediate window into the end-of-month window. Unlike Anthropic, the total probability is low (~13% combined), so this reads as the launch being pushed out rather than confirmed.

4. News-to-price correlation is weak here, which is the most important finding. The top 15 TechCrunch stories — Warp's AI software factory, OpenAI's teen ChatGPT, Stripe's reported $7B OpenRouter acquisition, Anthropic's $65B annualized revenue, Nvidia's $1.5B SoftBank data-center investment — contain no direct headline about an upcoming Anthropic Mythos-Class or Google Gemini 3.5 Pro launch. The biggest price moves are therefore driven by trading activity (likely a launch-date leak or an internal schedule) rather than the public news flow. This is a genuine price/news divergence: the market is pricing in an event the public headlines have not yet reported.

5. Two secondary divergences are worth flagging. On Polymarket, "Will Ari Weinstein leave OpenAI by Dec 31, 2026?" halved from 49% to 25% — a large de-risking move with zero matching news, suggesting an internal rumor that he is staying. And "AI wins IMO gold medal in 2026?" slid 28% to 22% with no relevant coverage; given the IMO timeline this may be a quiet re-rating of feasibility rather than news. The high-conviction anchors are stable: Anthropic IPO before 2027 sits at 86%, GPT-6 by Dec 31, 2026 at 82%, and claude-opus-5-max is the consensus top model for the month at 87% on Kalshi — all unchanged from the prior snapshot.