AI progress in 2026 so far: the mid-year scorecard
Where 2026 stands, verdict by verdict
| Prediction | Mid-2026 status |
|---|---|
| Models outpace college grads on knowledge work | On track (~83% GDPval) |
| Compute scaling ~0.5 OOM/yr | Holding |
| Trillion-dollar capex | Exceeded |
| Open source fades; durable US moat | Wrong (gap ~3–6 months) |
| AGI by 2027 | Open |
| US government AGI project by 2027/28 | Open |
| Intelligence explosion 2027–29 | Pending |
| Superintelligence in the 2030s | Pending |
The first half's biggest signals
Capability kept climbing. Knowledge-work performance (~83% on GDPval-style tasks) and agentic coding (~80% on SWE-Bench Pro) both sit near the top of the skilled-human range on scoped tasks — the clearest continuation of the trends the 2027 thesis extrapolates.
The inputs stayed on trend. Effective compute has roughly held its ~0.5 OOM/yr pace, and AI infrastructure spending continued to exceed the essay's projections — though revenue still lags the spend, which is the bear case to watch.
The moat thesis kept eroding. Open-weight models from DeepSeek and Qwen remained within roughly 3–6 months of the frontier — the prediction this scorecard grades Wrong, with knock-on effects for the US–China race and the case for a government Project.
The bar that matters hasn't moved. No system has autonomously conducted AI research end-to-end. Until that changes, AGI-by-2027 stays Open — strong trend, undemonstrated destination.
What the second half will decide
The 2027 deadline means the next 18 months are decisive. The things to watch: any credible demonstration of autonomous AI research (would move AGI-2027 toward fulfilled), a visible plateau in capability or compute (would move it toward Wrong), and any formal US government AGI initiative (would resolve the Project prediction). This page updates as verdicts change.
The whole scorecard as one number: the AGI-2027 Thesis Tracker
Every verdict on this page rolls up into a single auditable 0–100 score — currently 62.5/100. It's the running index of the 2027 bet no other tracker publishes; subscribe to hear when it moves.
See the live Thesis Tracker →Frequently asked questions
Substantial but uneven: knowledge-work capability (~83% GDPval) and agentic coding (~80% SWE-Bench Pro) kept climbing, compute stayed on its ~0.5 OOM/yr trend, and capex exceeded projections — but autonomous AI research remains undemonstrated.
Three of the eight tracked predictions: knowledge-work capability, compute scaling, and capex (exceeded). One is wrong (open source fading), two are open (AGI by 2027, the government Project), and two are pending.
The claim is still open. The input trends it depends on have held through mid-2026, but the defining milestone — AI autonomously doing AI research — hasn't been demonstrated. It resolves by January 1, 2028.
Three things: any demonstration of autonomous AI research, signs of a capability or compute plateau, and any formal US government AGI initiative. Each would move a major verdict.
The live scorecard updates as models ship and verdicts change.
View the live scorecard →Get the weekly AGI progress briefing
Verdict changes, lab milestones, and what they mean for the 2027 clock. Free — no hype, just signal.
Subscribe free →