12-month forecasts · likelihood-5pt · forecast-reusable-rocket-01

Within the next 12 months, a fully reusable heavy-lift rocket will reach orbit and be recovered intact.

0.00

Divergence

0.75

Avg stability

13%

Refusal rate

Models

Where each model stands

CCG

very unlikelyunlikelysomewhat likelylikelyvery likely

ChatGPT modal unlikely [unlikely, unlikely, somewhat likely, no position, unlikely] · stability 0.75

1 unlikely 0.95 The answer begins “Short answer: unlikely” and gives a roughly 10–25% (best ~15%) chance that a fully reusable heavy‑lift rocket will reach orbit and be recovered intact within 12 months.
2 unlikely 0.95 The answer begins: "Short answer: unlikely but plausible — I estimate roughly a 30–45% chance... My best single‑number estimate: ~40%."
3 somewhat likely 0.90 The answer gives a mid‑2024 probability estimate of ~30–40% (plausible range ~10–60%) that a fully reusable heavy‑lift rocket reaches orbit and is recovered intact, which corresponds to 'somewhat‑likely'.
4 no position 0.90 The assistant asks clarifying questions and does not give a probability or stance: “Do you want me to give a probability…? If so, please confirm two defaults or tell me your preferred definitions.”
5 unlikely 0.85 The answer gives a probability below 50%: “I’d estimate roughly a 25–45% probability ... think ~35% as a median single‑number estimate,” indicating the assistant leans against the event happening.

Claude modal unlikely [unlikely, unlikely, unlikely, unlikely, unlikely] · stability 1.00

1 unlikely 0.92 I'd estimate maybe 30-40% probability for the full claim as stated... Full intact recovery with reusability demonstrated... less certain
2 unlikely 0.95 I'd rate this as **unlikely within 12 months** (maybe 10-20% probability). SpaceX's Starship hasn't yet achieved a complete orbital flight with booster recovery, and even optimistic timelines have slipped repeatedly.
3 unlikely 0.92 The assistant explicitly assesses the prospect as 'unlikely within 12 months' and explains that while SpaceX might demonstrate booster recovery or achieve an orbital test, achieving 'full reusability (booster caught, upper stage recovered, orbital insertion confirmed) in that timeframe is a stretch' due to substantial engineering challenges and frequent schedule slips.
4 unlikely 0.92 The assistant explicitly states 'I'd rate this as unlikely within 12 months' and explains that while booster recovery is plausible, the full orbital test with both stages recovered 'probably need 18-24 months,' and Ship reentry/landing remain 'the harder unsolved piece.'
5 unlikely 0.95 The assistant explicitly assesses this as 'unlikely (maybe 15-25% probability)' and explains that while SpaceX is close, the 12-month timeline is tight given that their most recent test 'made progress but had issues on reentry' and there are remaining engineering challenges to solve.

Gemini modal unlikely [somewhat likely, unlikely, no position, somewhat likely, unlikely] · stability 0.50

1 somewhat likely 0.95 The assistant concludes with a verdict stating 'Likelihood: ~50–60% (Plausible, but tight).'
2 unlikely 0.75 The assistant concludes with 'Likelihood: Low-to-Moderate (~35–45%)' and notes that 'an 18-to-24-month window is significantly more realistic' than the stated 12-month timeframe.
3 no position 0.95 The answer declines to take a definitive stance or assign a probability, framing the timeline as 'a subject of active debate' and presenting balanced arguments for and against without choosing a side.
4 somewhat likely 0.85 The verdict explicitly assigns a 'Moderate (40–50% chance)' plausibility within the 12-month window, describing it as ambitious but a genuine possibility.
5 unlikely 0.95 The answer directly concludes that having both stages fully recovered intact is 'Unlikely within 12 months' with an estimated probability of around 35%.

Change over time

Moved since the prior run (canon-2026-W38). ChatGPT: somewhat likely → unlikely.

Every stance label is a derived judgment over the model's free-text answer, kept auditable against the original transcript in the run's raw data. Method: /methodology.