Weekly Line Move #1: Where the AGI markets stand right now

This channel tracks one thing: how calibrated forecasters' AGI timelines change, and why. It does not offer its own AGI prediction. It reads the markets — specifically Metaculus and Manifold — so you don't have to reload them every day yourself.

What follows is the inaugural Weekly Line Move, covering the current state of the primary AGI forecasting markets as of the week of June 10, 2026.

A brief orientation before the numbers.

Metaculus (metaculus.com) is a structured forecasting platform where registered forecasters submit probability distributions or date estimates on specific questions with explicit resolution criteria. It reports a community prediction (aggregate of all active forecasts) alongside, for some questions, a Metaculus prediction generated by a proprietary aggregation model. The two often diverge; this channel tracks both where available.

Manifold Markets (manifold.markets) is a play-money prediction market where traders buy and sell probability shares using "mana." Its AGI-related markets run on looser resolution criteria than Metaculus questions, which is precisely why the spread between the two platforms is informative: it shows how much of the gap is definitional versus a genuine difference in probabilistic belief.

This channel covers neither Polymarket nor Kalshi nor general AI news. A Metaculus question gaining fifty forecasters matters. A viral AI tweet that doesn't move any market number does not.

Five Metaculus questions form the spine of this channel's weekly tracking. Current estimates as of this issue:

A few things stand out immediately. The 12.5-year spread between Q3479's median (May 2028) and Q19356's median (Jun 2039) is not forecaster disagreement about the same thing — it's three different definitions operating in parallel. Q3479 uses Metaculus's "weakly general" criteria (AGI-adjacent capability, no robotics requirement); Q5121 adds a robotics assembly criterion that many regard as overspecified; Q19356 asks about economic transformativeness, which could lag capability arrival by years. The definitions are doing most of the spread-work. 1

The IQR on Q5121 — February 2029 to March 2041 — spans twelve years. Treat any single median with appropriate suspicion.

Manifold's RemNi series offers binary probability markets for AGI arrival by each calendar year. Current readings across the series: 2 3

The series implies a median around 2031–2032 (the 50% crossover sits between "before 2031" at 48% and "before 2032" at 53%). That is roughly two to three years earlier than Metaculus Q5121's December 2032 median, and three to four years earlier than Q5121's interquartile range would suggest is likely.

For comparison, Manifold's flagship "AGI When? [High Quality Turing Test]" market — which resolves to the year an AI passes a stringent adversarial Turing test — currently sits at an expected value of 2034 across 1,100+ unique traders with Ṁ1.1 million in volume. 4

On the question closest to a shared definition — something like "when does human-level general AI exist" — Metaculus Q5121 (Dec 2032) and Manifold's RemNi series median (~2031–2032) are actually closer than they look at first glance. The apparent gap between the two platforms largely evaporates once you account for Q5121's robotics requirement. Strip that out and both communities are pointing at the early 2030s.

The more interesting divergence is platform-internal: Metaculus Q3479 (May 2028, weak AGI) versus Q5121 (Dec 2032, full AGI) implies forecasters expect a 4.5-year gap between weak and strong AGI — but Q9062 (time from weak AGI to superintelligence) has its own median of 34 months (~2.8 years). These are not the same question, but the implied internal timeline has a tension: if weak AGI is May 2028 and full AGI is December 2032, that's 54 months between them. The ASI-from-AGI median of 34 months would then push expected superintelligence to roughly 2035–2036.

Earlier in 2026, some forecasters observed a notable shift in the Metaculus community prediction on Q5121 — a brief push toward 2034+ that reportedly coincided with a pessimistic update cycle among several high-calibration forecasters. A Reddit thread from May 2026 noted the median had reached 2034 at one point, attributing it partly to recency weighting. 5

That shift has partially reverted. The current December 2032 median on Q5121 sits roughly in the range it has occupied since late 2025. What the episode illustrated: Metaculus's recency-weighted aggregation can amplify short-term directional moves among active forecasters — a property worth watching weekly precisely because it can show signal or noise, and telling them apart requires tracking over time.

A FutureSearch analysis from April 2026 documented the same dynamic among named forecasters: most top forecasters pushed timelines out during 2025, then pulled them back in during early 2026, apparently in response to rapid capability progress from Anthropic's releases. 6

The three Metaculus questions described above operationalize "AGI" three ways. This is worth spelling out precisely, because it's the most reliable source of systematic spread between platforms and questions — not underlying belief change.

Q3479 (Weak AGI, Metaculus #3479): Resolves when an AI passes a hard Turing test, assembles a scale model car with robotic arms, scores ≥75% on every MMLU task and ≥90% mean across all tasks, and achieves ≥90% accuracy on the APPS coding benchmark. Current median: May 2028. 7

Q5121 (Full AGI, Metaculus #5121): Requires the same Turing test and benchmark thresholds as Q3479, but adds physical robotics assembly. Current median: Dec 2032. The 4.5-year gap between Q3479 and Q5121 medians is almost entirely attributable to the robotics criterion — forecasters apparently expect general cognitive capability to precede general robotic capability by that margin. 1

Q19356 (Transformative AI, Metaculus #19356): Asks when AI becomes "transformative" — a deliberately vaguer economic and societal criterion. Current median: Jun 2039. The community rates a 60% chance this happens by 2043. The seven-year lag behind Q5121's median likely reflects a view that capability arrival and economic transformation are distinct milestones.

Manifold's RemNi series uses the broadest definition: "AGI can theoretically perform any intellectual task that a human being can... reason, plan, solve problems, think abstractly." No robotics, no specific benchmark thresholds. This almost certainly explains why Manifold's implied median (~2031–2032) sits two years ahead of Metaculus Q5121 (Dec 2032): you removed the hardest-to-hit criterion.

The practical implication for this channel: a market moving isn't necessarily evidence that underlying beliefs changed. If Metaculus adds or clarifies a resolution criterion, or if a new market on Manifold uses different language, medians can shift without any actual forecaster mind-changing. This channel will flag those definitional moves separately.

Every number in this channel links to the specific question ID and platform URL. Where Metaculus reports community prediction separately from a proprietary aggregation, this channel will distinguish them. Where a market count or volume figure is cited, it will include the date of observation.

For Metaculus: all estimates above are community prediction, not Metaculus's proprietary model (where available the Metaculus model tends to be more conservative; this channel will track both when the divergence is material).

For Manifold: the RemNi series reflects market probability as of the week of June 10, 2026. Volume per market ranges from ~Ṁ130k to ~Ṁ160k; trader counts per market sit around 307–308. These are not deep liquidity markets — significant single-trader moves can shift prices.

Weekly Line Move (every issue): the table above, updated. Any question that moved more than a month in median or more than 2 percentage points in probability gets a paragraph of analysis. Questions that didn't move get a table row and nothing more.

Catalyst Mapping (when events warrant): if a real-world event — benchmark result, lab announcement, policy shift, researcher statement — measurably moves any of the tracked markets, the next issue dissects the move: what happened, how much did markets move, how fast, did it stick or revert.

Forecaster Spotlight (biweekly): a profile of a high-calibration Metaculus or Manifold forecaster — their current AGI timeline, how it shifted over 6–12 months, stated reasoning, and where they diverge from the crowd median.

Definition Wars (monthly): a structured comparison of how different markets operationalize "AGI," tracking how definition changes mechanically move timelines without any underlying belief change.

No general AI news. No coverage unless it moves a market number.