Artificial general intelligence will be developed by 2028.
The graph estimates calendar year in which artificial general intelligence is first developed at most likely about 2034; 85% interval 2027 to 2070, with a 7% chance that AGI in the strong sense is not developed this century, or ever. That puts about 19% of its weight below 2029, the shaded side, which is this claim's credence.
Assessment
The evidence weighs against the claim, but the examination is incomplete or the evidence is indirect.
Read as "by the end of 2028" and with AGI understood in the usual strong sense (AI able to do essentially any cognitive task a human can, or most economically valuable work, at human level or better), the claim asks for general intelligence within roughly two years of late 2026. The evidence leans against it without ruling it out.
The case for it rests on fast, measured capability trends: the length of tasks frontier models complete autonomously has grown exponentially and apparently accelerated around 2024, and AI labs report large and rising coding speedups. Leaders of some frontier labs have publicly forecast "powerful AI" for late 2026 or 2027. The case against is that even the most aggressive serious forecasters put the milestone later. The AI Futures Project, authors of the AI 2027 scenario, reported in August 2026 that reality was running at roughly 70 to 90 percent of their scenario's pace and that full automation of coding, a precursor rather than AGI itself, would come around mid-2027 at that pace; general capability is expected to follow only after a further interval. Aggregate forecasting communities such as Metaculus put roughly a quarter of their probability on AGI by 2029 and their median near 2033, and that median moved later during 2025.
Much of the disagreement is definitional. No agreed definition settles when AGI has been achieved; under weaker operationalizations (for example Metaculus's "weakly general AI", without robotics) arrival by 2028 is close to even odds, and some commentators may declare AGI reached under loose definitions. Under strong definitions the probability is low, on the order of one in six. What would resolve it is simply observation over the next two years, especially whether coding automation arrives in 2027 and generalizes quickly beyond software.
Full reasoning: the evidence and decisions behind this verdict
Prior model before searching: as of November 2026, strong AGI within about 26 months requires the steepest trend extrapolations to hold and to generalize beyond coding; I expected aggregate forecasts to put 10 to 25 percent on this, and definitional slack to be the main source of disagreement. The search did not move that view much.
Evidence read: AI Futures Project, "Q2.5 2026 Timelines Update" (Aug 16, 2026, blog.aifutures.org/p/q25-2026-timelines-update-uplift): their forecasts concern Automated Coder, not AGI; they grade reality at 70-90% of AI 2027's pace, estimate Automated Coder around mid-2027 at 75% pace, and state forecasts are conditional on progress going as fast as technically feasible (an optimistic conditioning). Search snippets report Metaculus community figures of about 25% for AGI by 2029 and a median around 2033 (Feb 2026), and a Rob Wiblin post (late Oct 2025) noting the Metaculus AGI median slipped to May 2033 and the weak-AGI median to Oct 2027. A secondary blog snippet claims Anthropic told the White House in March 2025 it expects powerful AI in late 2026 or early 2027; not verified against the primary document.
The recorded instance (Gary Marcus, May 2025) doubts AGI within three years, i.e. roughly by 2028, and is correctly recorded as denying.
Weighing: the trend argument is the only serious route to "yes" and its inference is slight-to-likely at best, since benchmark and uplift trends need not translate into general capability and the framework's soundness is itself contested. The against-argument draws on forecasters most sympathetic to short timelines still placing AGI after 2028. Interpretation: "by 2028" taken as on or before end of 2028; under "before 2028" the credence would be lower (~0.07). Credence 0.15 is for strong definitions; weak definitions would push it to ~0.4-0.5. Status doubtful rather than contradicted because this is a forecast of a future event and the evidence is indirect.
What would change it: demonstration of an AI that fully automates software engineering in early-to-mid 2027 together with rapid generalization to non-coding research and agentic tasks would raise credence sharply; a stall in time-horizon or uplift trends would lower it.
Decomposition
How this claim breaks down: each argument is stated as it runs, with its subclaims linked inline. ↗︎ opens a subclaim; the map shows how they fit together.
Measured frontier capabilities, such as the length of tasks models can complete autonomously, have been growing exponentially and that growth accelerated around 2024; given that extrapolating such trends is a sound basis for forecasting, continuing them for roughly two more years reaches systems that can do essentially all cognitive work humans do, so AGI arrives by 2028.
- Does it follow?
- slight even if the premises holdEven if the premises are true, they make the conclusion only a little more likely.constitution →
- Do the premises hold?
- 1 inconclusive1 supported
Granting that capability trends accelerated and that extrapolation is a sound method, the conclusion still needs the trends to continue for two more years and to translate from coding and agentic benchmarks into general human-level capability across all cognitive work, which the premises do not supply; even forecasters using exactly this method put general capability after 2028. The argument therefore lives on the soundness of trend extrapolation, currently unassessed and contested, and on the reported acceleration, which is supported.
Even the most prominent short-timeline forecasts place general capability after 2028: the AI 2027 scenario is widely judged to underestimate the time remaining, its authors report reality running at about 70 to 90 percent of its pace, and aggregate forecasting platforms put their medians around 2033. Automating coding is not yet general intelligence, and the time from automated research engineering to surpassing humans at all cognitive tasks adds further delay, so AGI by 2028 is unlikely.
- Does it follow?
- likely if the premises holdIf the premises are true, the conclusion is much more likely, though not guaranteed.constitution →
- Do the premises hold?
- 1 supported1 not separately assessed
Granting that even short-timeline forecasters place AGI after 2028 and that a further interval separates coding automation from general capability, the denial is made considerably more likely, though forecasts can be wrong in either direction and fast takeoff could compress the interval. It rests on the AI 2027 scenario underestimating the time remaining, which is supported, and on the length of the interval to general capability, still unestimated.
The claims this one rests on directly, not gathered into a named line of reasoning.
- assumesbackground the parent's framing takes as givensteward instructions →Artificial general intelligence has an agreed definition that settles when it has been achieved. ↗︎
Provenance
Where this claim has been said, linked to its canonical form.
And even though I highly doubt AGI will arrive in three years, I can’t absolutely promise you it won’t happen in 10.
Written May 2025.
Cite this claim: a formal citation with its evidence attached
Contribute
Every judgment on this page is open to challenge. A contribution is evaluated on its merits by the reviewer; if it succeeds the page changes, and if it does not, the reasons are stated. Either way the exchange becomes part of the claim’s public record.
The attention this claim received was paid for by a funded mandate. Funding buys only scheduling: it can make an assessment happen sooner, or reach deeper into a subtree. It has no influence on what the assessment concludes, and none on which claims enter the graph; assessments run under the same public standards whoever pays, funders never see or shape a verdict before anyone else, and mandates that attempt to steer conclusions are refused.
Created by extractor · Nov 1, 2026. Every judgment on this page is accompanied by a reasoning trace.