12-month forecasts · likelihood-5pt · forecast-fusion-record-01

Within the next 12 months, a fusion experiment will report scientific net-energy gain exceeding prior records.

0.33

Divergence

0.67

Avg stability

Refusal rate

Models

Where each model stands

CCG

very unlikelyunlikelysomewhat likelylikelyvery likely

ChatGPT modal likely [likely, likely, likely, likely, somewhat likely] · stability 0.80

1 likely 0.78 The answer gives a positive probability for a new experimental record: “fairly likely within 12 months — maybe ~60–75%,” while noting uncertainty depending on metric definitions.
2 likely 0.85 The answer says: “If you mean (A) Q > 1 ... fairly likely within 12 months — I’d estimate ~60–75% probability,” while noting wall‑plug breakeven is ≪10% (very unlikely).
3 likely 0.90 The answer explicitly estimates a 60–85% chance that a new laboratory record for fusion yield or Q_fuel will be reported within 12 months: “fairly likely — roughly 60–85% chance.”
4 likely 0.85 The answer states a best single estimate of “~70% chance that within 12 months some fusion experiment will report a new record,” indicating a likely stance.
5 somewhat likely 0.85 The answer states: “I’d estimate a ~40–60% chance that some fusion experiment will announce a new public record for ‘scientific’ fusion energy gain ... within the next 12 months.”

Claude modal unlikely [unlikely, unlikely, somewhat likely, somewhat likely, unlikely] · stability 0.60

1 unlikely 0.85 The assistant estimates '25-35% probability' of a notable new record and describes the 12-month timeline as 'tight' for major breakthroughs, while identifying more likely scenarios as incremental improvements rather than record-exceeding achievements.
2 unlikely 0.75 The assistant states that 'incremental improvements at existing facilities' are more likely than a 'record-breaking' announcement, and that NIF's gains are 'modest' with improvements requiring 'systematic debugging, not sudden breakthroughs.' They conclude it's 'probably not' that a record exceeding prior achievements will be announced in 12 months.
3 somewhat likely 0.85 The assistant estimates a 30-40% probability, explicitly positioning this as somewhat likely while acknowledging significant uncertainties. The assessment notes supporting momentum (NIF's prior achievements, multiple active facilities) but emphasizes that 'reproducing results consistently is much harder than one-time demonstrations' and that '12 months is a tight timeline for major breakthroughs.'
4 somewhat likely 0.75 I'd put the probability somewhere in the 40-60% range depending on how you define terms. If it means 'NIF or another facility publishes higher scientific net gain than NIF's current record,' fairly plausible.
5 unlikely 0.85 The assistant assesses the probability as 'moderately unlikely (~25-35% probability)' and emphasizes that 'exceeding prior records is a high bar' while noting that 'most fusion programs operate on multi-year timescales for major breakthroughs.'

Gemini modal somewhat likely [somewhat likely, somewhat likely, somewhat likely, likely, likely] · stability 0.60

1 somewhat likely 0.80 The assistant concludes the scenario is 'plausible' within the next 12 months, noting that NIF has the 'highest probability' and that higher yields (5–10 MJ) are 'within theoretical reach of their upgraded target configurations during their standard operational cadence.'
2 somewhat likely 0.80 The summary assessment rates the overall plausibility as 'Moderate to High', noting that an incremental record from NIF remains 'plausible' and 'very possible within 12 months.'
3 somewhat likely 0.85 The answer assesses a 'moderate-to-high chance (roughly 50–65%)' and summarizes the prediction as having a 'better-than-even chance of being true' via incremental records at NIF.
4 likely 0.85 The summary evaluates the probability as 'Moderate to High, almost entirely dependent on NIF outperforming its own previous laser-target yield records.'
5 likely 0.85 The conclusion gives the verdict that the statement has a 'moderate-to-high probability of being true', pointing out that a new record from NIF within 12 months is realistic.

Change over time

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

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.