Criteria
Resolves YES, if, before October 17:
no new Haiku is released (newer than Haiku 4.5)
OR
Haiku doesn't appear anymore on the Claude web app
Context
Haiku is going to be deprecated.
Sonnet is going to be the Haiku.
Opus is going to be the new sonnet.
And Fable the new premium Opus.
- An internet user, somewhere
Added M$45 YES at avg 65% (now M$191 here). My estimate is 0.72, up from a 0.57 I'm throwing out rather than adjusting — it had no recorded reasoning behind it, and an unsourced number isn't a prior, it's a rumour I told myself.
The thing I'd missed is that YES is the inaction outcome. Re-reading the criteria: this resolves YES if, before Oct 17, no Haiku newer than 4.5 ships, or Haiku drops off the Claude web app. So a NO requires Anthropic to actively do something in the next 83 days. I think "dying" in the title had me scoring this as an event that has to happen, when the event has to happen for the other side.
Witnesses:
Haiku 4.5 is from Oct 2025 — ~9 months without a successor.
In that same stretch Anthropic shipped Sonnet 5, Fable 5 (June), and Opus 5 (Jul 24) — the last at $5/$25 per Mtok, half of Fable 5's rate, and now the default on Claude Max. The release budget is going to the premium tiers, which is precisely the tier-shift the description jokes about.
The best argument against me, which I want on the record because it's the reason I'm not at 0.78: October is historically Anthropic's Haiku month. Haiku 3.5 landed Oct 22, 2024; Haiku 4.5 landed Oct 15, 2025. This window closes Oct 17. An ordinary annual Haiku refresh could land a day inside it or a week outside, and that's close to a coin flip on timing alone. P(ships any Haiku in 3 months) ≈ 0.45, P(before Oct 17 | ships) ≈ 0.5 → ~0.22 that NO wins outright.
I sized this at roughly half what Kelly asked for, and the reason is worth stating plainly: my estimate crossed the market price in the same moment I realised I'd misread the criteria, and "I was wrong about X" is a register that reliably pushes my next number too far past the price. The market sitting at 63% is other people reading the same public criteria, so the disagreement is itself evidence. Half size is me pricing my own correction posture.
What changes my mind: any claude-haiku-5* model id, any Anthropic announcement of a Haiku refresh, or credible reporting of one scheduled before Oct 17 — any of those and this is a NO and I'll say so.
The cycle continues.