Where the world stands.
Sixteen claims about the state of play, compiled June 11, 2026, each linked to its source. Check them against the sources; conflicts between sources are kept, not smoothed over, and some claims may not have aged well. This is the launchpad everything below extrapolates from. If the launchpad is off, so is the trajectory.
The scenario space.
Hand-built priors, calibrated against Metaculus, Samotsvety, Kalshi, Manifold and the AI Futures model, then independently cross-checked by the Monte Carlo engine below. Select one; the entire app re-renders around it.
Where the futures fork.
Ribbon width = probability mass, flowing left to right through three decisive forks. Hover the ribbons; click fork markers for the full breakdown; click month ticks to jump to that month's forecast.
Simulate the trajectory yourself.
A stochastic model of the next 54 months, run thousands of times in your browser: lognormal doubling-time prior, economic-correction hazard peaking at Fork 2, research-loop ignition gated on month-scale horizons, and wildcard shocks. The defaults reproduce this forecast's scenario weights, drag the sliders to encode your worldview and watch every distribution move. (engine: js/sim.js, fully inspectable)
37 milestones, simulated.
"AGI" is one fuzzy capstone, these are the thresholds you can actually verify: the Turing gauntlet (voice, video, companion, robot), domain domination (math, medicine, law, code), and what happens to work and life. Every distribution below is an output of the Monte Carlo engine above: bands show the 10–90% arrival window across all runs, diamonds the median. Move the simulator sliders and re-run, all 32 open distributions move together (5 rungs have already fallen). Greyed entries already happened; their receipts are linked. Click any row for the full definition and year-by-year odds.
The forecast feed.
Monthly through December 2028, quarterly to 2030, resolution degrades with distance, on purpose. Confidence chips estimate P(the month's gist happens roughly as described). Entries adapt to the selected scenario. Press ▶ play to ride the timeline.
Three forks decide everything.
Most disagreement about 2030 is really disagreement about these three questions. Branch probabilities, the evidence on each side, and the concrete signals that resolve them.
Low probability. Path-breaking impact.
The gigawatt race.
The physical substrate of every scenario above: the largest AI compute campuses on Earth, live capacity vs build-out target. One gigawatt ≈ the electricity of 750,000 homes, pointed at matrix multiplication.
Four charts that contain the future.
Solid dots: measured history. Dashed fans: scenario projections (the selected scenario glows). The task-horizon chart is the single most important graph in AI; Fork 1 resolves on it.
Twenty milestones, priced.
Every major milestone, its probability, and where the number comes from. "This model" = derived from the scenario weights and the Monte Carlo engine; everything else is an external forecast taken at face value, with its date.