AGI, fusion, quantum advantage, longevity breakthroughs — markets are structurally bad at pricing events with no historical precedent, because "no precedent" is exactly the condition under which the human mind defaults to binary yes/no thinking instead of a probability distribution. This letter is about the framework, not the forecast.
Most investment analysis implicitly assumes the future looks like a smoother version of the past. AGI, fusion, quantum computing, and longevity breakthroughs break that assumption directly — there is no historical base rate for "humanity builds a second kind of general intelligence" to extrapolate from. This letter is about how to think about that honestly, rather than either dismissing these events as science fiction or treating any single date as a certainty.
Metaculus, a forecasting platform with nearly 2,000 contributors and a genuine track record on near-term political and economic questions, has run a standing question on AGI arrival for years. As recently as 2020, the community's median estimate was roughly fifty years away. By February 2026, that same community's aggregate forecast had compressed to a 25% probability of AGI by 2029 and 50% by 2033. Samotsvety Forecasting, a superforecaster group with a verified competitive track record on platforms like INFER, moved from a 32% probability of AGI by roughly 2042 (their 2022 estimate) to roughly 28% by 2030 (their January 2026 update) — more than a decade of timeline compression inside three years of calendar time, from people whose job is calibrated probability, not hype.
The honest caveat this letter will not soften: forecasters disagree sharply on definitions, and the compression itself could partly reflect excitement rather than genuine evidence. Yann LeCun and Gary Marcus, credentialed AI researchers, argue current architectures cannot reach AGI at all under any timeline. Lab CEOs (Dario Amodei, Sam Altman) predict systems broadly better than humans at almost everything within one to two years — a group selected, by the nature of their jobs, to be bullish on their own technology. The right response to this spread is not to average it into a false consensus. It is to hold a genuinely wide probability distribution and size any related position accordingly.
Fusion energy has spent fifty years as a punchline — "always thirty years away." That joke has stopped being accurate without becoming a confident timeline. Commonwealth Fusion Systems' SPARC demonstration tokamak is roughly 75% complete at its Devens, Massachusetts site, targeting first plasma in late 2027, and has secured power-purchase agreements with Google and Eni for a planned 400-megawatt commercial follow-on plant. Over $11.4 billion in disclosed private capital has flowed into commercial fusion companies as of mid-2026. Against that, ITER — the largest fusion project in human history, a 35-country collaboration — has slipped its own first-plasma date to the mid-2030s and full deuterium-tritium operation to 2039, with costs now exceeding $22 billion. Private, compact-tokamak fusion (Commonwealth, Helion, Tokamak Energy) is on a visibly faster track than the public megaproject that originally defined the field's timeline — a genuine, trackable divergence, not a rounding error.
The mistake this letter is naming directly: treating whether AGI, fusion, or quantum advantage arrives as a binary bet rather than as a probability distribution with a wide confidence interval that should update, gradually, on new evidence. A more disciplined approach borrows directly from how options are priced — not whether an event happens, but the expected value across a distribution of dates and outcomes, and what holding optionality on that distribution costs today. Concretely: back the infrastructure and picks-and-shovels layer that pays off across most scenarios (grid capacity, superconducting materials, advanced compute) rather than concentrating a bet on a single named breakthrough or company; treat published forecaster timelines (Metaculus, Samotsvety) as a genuine input, not noise, precisely because their track record on calibration is better than most individual experts'; and re-underwrite the position on a fixed schedule as milestones resolve (SPARC's first-plasma result in 2027 is a real, dated, binary event this letter can revisit directly) rather than holding a static thesis indefinitely.
The right question is never whether this happens. It is what the probability distribution looks like today, what it costs to hold exposure across that distribution, and what specific evidence would genuinely move it before this letter revisits the question next.
Superforecaster timelines on AGI have compressed by roughly a decade in three years of calendar time, and private fusion capital ($11.4 billion and rising) is now visibly outpacing the public megaproject that used to anchor the field's entire timeline. Neither fact license a confident date. Both license taking the underlying uncertainty seriously enough to hold real, sized exposure to the infrastructure layer that pays off across a wide range of outcomes, rather than either dismissing these technologies as science fiction or making a binary bet on a specific year. The risk this framework is most exposed to is definitional drift — as this letter's own sources note, the goalposts for "AGI" keep moving precisely because each new capability milestone gets redefined as insufficient once achieved, which means the actual resolution of any of these questions may be far messier than the clean forecast dates suggest.
Founder, NextGen Economics · Bangalore, India · July 2026
Sources: Metaculus community AGI forecast (Feb 2026 update) · Samotsvety Forecasting AGI updates (Jan 2026, 8 forecasters) · 80,000 Hours (review of expert AGI forecasts) · Fusion Industry Association (private fusion investment tracker, June 2026) · Commonwealth Fusion Systems & ITER public project timelines · Congressional Research Service, "Toward Commercial Fusion Energy" (Feb 2026).
Not investment advice. This letter proposes a framework for reasoning about deep uncertainty; it does not constitute a recommendation regarding any security or a forecast of any specific outcome.