Quebec adds a correlated risk. The Parti Québécois leads polls ahead of the October 2026 provincial election, though sovereignty sentiment in Quebec is at historic lows. The Base Rate Analyst dismissed a pre-2027 Quebec referendum as improbable, assigning it roughly 5-10%, but acknowledged that a PQ victory would change the calculations for Alberta's political timing.

The Insider sees Smith stalling: "She has no incentive to lose a referendum. Her leverage comes from keeping the threat alive rather than resolving it. The petition creates an obligation, but an obligation can be slow-walked." The Contrarian pushed back, arguing that the petition's success, combined with the Trump administration's meetings with Alberta separatists, has generated momentum that may prove difficult to redirect.

The Insider leans toward a shorter delay: "The model exists. The question is when they're ready to announce. The $380 billion valuation from February demands a 'generational leap' narrative, and incremental 4.x updates don't deliver that."

The Systems Thinker noted that Anthropic's release cadence has accelerated dramatically over the past 18 months, from 6-14 months between major versions to something closer to continuous iteration. At the same time, the company has historically exhausted a version line before jumping, and 4.6 left room for 4.7.

The Base Rate Analyst was the most sceptical, assigning just 15% probability: "Past performance predicts future behaviour. The February rumours were wrong. There's no announced event. The version numbering suggests more 4.x runway." The Insider pushed back: "Google Cloud Next is April 22nd. That's your announcement window."

We're more sceptical than the market because of the May 6th "Code with Claude" event.

Mythos is Anthropic's forthcoming "agentic" tier, described in leaked reports as having a "step change in cyber capabilities." Early access testing began in late March, per Fortune, but that's a tight window for a public release. The Contrarian assigned just 25% probability, arguing that "Anthropic's safety-first brand makes rushing a model described as potentially 'too dangerous to release' strategically incoherent. The May event exists for a reason."

The Scenario Planner was the outlier at 69%, citing competitive pressure from OpenAI and Google as forcing functions that could accelerate the timeline. The Risk Analyst split the difference: "Safety reviews take the time they take. But if they've been running since March, April 30th becomes difficult rather than impossible."

The disagreement hinges on what "release" means. If the question covers expanded early access rather than general availability, probabilities shift upward. If it requires a full public launch, the May 6th event becomes the natural venue, and April becomes a near-miss.

Our estimate: 22% chance Meloni is out as Prime Minister by June 30th. Polymarket prices this at just 8%. The gap is meaningful, though the direction favours continuity either way.

Meloni's government is already the third-longest lasting since 1946, which says something about both her political skill and the chaos of Italian politics. The March referendum defeat wounded her, but it didn't collapse the coalition. The Sceptic noted that "Italian governments rarely collapse slowly. They collapse suddenly, or they persist. The question is whether the referendum defeat, combined with economic headwinds, creates the spark."

The Insider is more relaxed: "The opposition is fragmented. PD and M5S can't form a credible alternative. Where does the replacement government come from?" The Risk Analyst pointed to economic stress as the wildcard, citing Meloni's own warning that 2026 would be "much worse" economically. Bond market pressure or a banking crisis could force a collapse, but those are tail scenarios.

Our estimate: 16% chance X is banned in any European country by December 31st 2026. Polymarket prices this at 22%. The difference is modest but directionally interesting.

The Digital Services Act structures enforcement as a ladder: fines, compliance orders, and legal challenges before any suspension. A full ban sits at the end of that sequence. No EU member state has ever banned a major social media platform, and the DSA explicitly prohibits arbitrary blocking. The Systems Thinker put it plainly: "The system is designed to escalate slowly. Bans require extreme provocation and exhaustion of alternatives."

X paid the €120 million fine, filed an appeal, and made marginal compliance adjustments. That's enough to avoid outright defiance without conceding the regulatory battle. The Insider noted that "a ban would trigger a diplomatic crisis with the Trump administration that European governments have no appetite for." The Contrarian pushed the probability higher, arguing that geopolitical tensions and Musk's role in U.S. politics could trigger emergency measures, but acknowledged this was a tail scenario.

Our estimate: 12% chance any stolen KitKats are recovered by April 5th. Polymarket sits at 15%. The gap is trivial; both markets agree this is a long shot.

The theft was sophisticated: 413,793 KitKat bars, taken via impersonation of law enforcement, with no truck or cargo found six days later. The critical 48-72 hour recovery window passed without progress. The Risk Analyst noted that "chocolate has no serial numbers, no traceability once repackaged. The grey market absorbs this kind of product within days, especially with Easter demand."

The Scenario Planner allowed for a 17% chance of a lucky break, perhaps the truck is found abandoned, or an informant tips off authorities, but acknowledged this was optimistic: "The sophistication of the theft implies pre-arranged distribution. You don't impersonate police and steal half a million chocolate bars without knowing where they're going."

The Contrarian was the most sceptical at 7%, arguing that the absence of any updates six days after the theft suggests the trail has gone cold entirely. The product is likely already dispersed across multiple countries.