← Live scorecard
Analysis

How close are we to AGI? A capability check

Last updated: August 29, 2026 · Updated as verdicts change
By the AGI Scorecard team · methodology & independence
Closer than skeptics say, further than optimists claim. As of mid-2026, frontier AI clears many knowledge-work bars — ~83% on GDPval, ~80% on SWE-Bench Pro, agents in production — but has not crossed the defining line: autonomously doing AI research end-to-end. So on capability we’re deep into “very capable assistant,” not yet “drop-in researcher.”
The part a chat answer goes stale on
62.5/100 how much of the AGI-2027 thesis is still standing — recomputed from eight graded predictions, as of 2026-08-08
See all eight verdicts, each with its pre-registered flip condition →

What's already here

The measurable state of play is strong. Frontier models perform at roughly college-graduate level on knowledge work (~83% GDPval), handle complex agentic coding (~80% SWE-Bench Pro), and run as agents in real workflows. On the capability curve, the trajectory Aschenbrenner described is On track.

The missing piece

The gap between "very capable assistant" and "AGI" is autonomy. The defining bar — a system that can autonomously conduct AI research end-to-end, the "drop-in remote worker" — has not been demonstrated. That single gap is why the headline AGI-2027 prediction is graded Open, not On track.

Where the forecasters land

ForecasterAGI timeline
Elon MuskBy end of 2026
Aschenbrenner2027
Hassabis~50% by 2030
Metaculus50% by 2033
Academic survey50% by 2047

That’s where the forecasters land. How close do you think it is?

2025–2620272028–302030s2040+ / never

One tap — see which real forecaster you side with, instantly. No sign-up.

The bottom line: measurably closer than a few years ago — the whole forecast distribution has compressed from ~2060 to ~2033 in six years — but the finish line (autonomous research) is unproven. The nearest concrete test, Aschenbrenner's 2027, resolves by January 2028. As of July 2026, our auditable AGI-2027 Thesis Tracker puts the whole thesis at 62.5/100 — the single running measure of how close we actually are, and it moves only when a verdict changes. See the live score →

"How close" changes only when a verdict flips — and each flip condition is already published.

Every number above has a pre-registered condition that would move it. The next scheduled test is the AGI-2027 claim itself, resolving January 1, 2028. Subscribe and you hear the day any verdict — and the 62.5/100 score — actually moves, not when a headline does.

Alert me when the score moves →

How close we are is also a portfolio question.

The evidence above splits cleanly: capability is close, autonomy is not. The market prices the same split — buildout names get paid either way, AGI names need the last step to land. One tap scores either side against the same eight verdicts.

Paid either wayNeeds the last step

Instant result, no sign-up. Not investment advice.

Frequently asked questions

How close are we to AGI?

Close on capability, not on autonomy. As of mid-2026 frontier AI reaches ~83% on GDPval and ~80% on SWE-Bench Pro, but no system autonomously conducts AI research end-to-end — the defining AGI bar.

What's still missing for AGI?

Autonomy. The gap between a very capable assistant and AGI is a system that can independently do AI research and own tasks end-to-end without human cleanup — undemonstrated as of mid-2026.

When will we reach AGI?

No consensus: forecasts range from 2026 (Musk) and 2027 (Aschenbrenner) to 2030 (Hassabis), 2033 (Metaculus), and 2047 (academic survey median). Expert medians have compressed from ~2060 to ~2033 in six years.

The live scorecard updates as models ship and verdicts change.

View the live scorecard →

Be first to know when a verdict flips

One email when one of the eight verdicts changes and the AGI-2027 Thesis Tracker moves — the single auditable score no other tracker has. Not a weekly newsletter; nothing arrives until the score actually moves.

Subscribe free →