Is AGI just hype? What a two-year scorecard actually shows

The case that it's NOT just hype

Several concrete, checkable bets have held up. Effective compute has kept scaling at roughly 0.5 orders of magnitude per year (Epoch AI). Capital expenditure on AI infrastructure has exceeded even aggressive projections. And frontier models now score around 83% on knowledge-work benchmarks like GDPval, with agentic coding near 80% on SWE-Bench Pro. These aren't vibes — they're measurable trends moving the way the optimists said they would.

The case that it IS overhyped

But the skeptics have real points too, and the scorecard flags them honestly. The prediction that open source would fade and proprietary labs would hold a durable moat is clearly wrong — open-weight models trail by only ~3–6 months. AI revenue is tracking well behind the run-rate the hype implied. The promised “shocking qualitative leap” hasn't landed as a felt discontinuity. And the defining bar — AI that autonomously does AI research — remains undemonstrated.

How to tell substance from hype

The trick is to stop arguing about “AGI” in the abstract and grade specific falsifiable claims. That's what this scorecard does: eight predictions, each with a verdict (On track / Wrong / Open / Pending) and a pre-registered condition that would flip it. As of mid-2026: 3 on track, 1 wrong, 2 open, 2 too early. That distribution — mostly-right on trends, wrong on diffusion, undecided on the finish line — is the actual signal under the noise.

The bottom line

Is AGI just hype? No — the underlying trends are real and measurable. Is it here, or a sure thing by 2027? Also no — the decisive capability is unproven and at least one core claim already failed. The mature read is that “hype vs. real” is the wrong binary; the right move is to watch the specific claims resolve. The biggest one, AGI by 2027, is settled by January 2028.

Frequently asked questions

No, but it isn't imminent either. Grading specific predictions two years on: compute, capex, and capability trends are on track; the claim that open source would fade is wrong; and 'AGI by 2027' is genuinely open, resolving by January 2028.

In parts. AI revenue is behind the hype's run-rate, the 'shocking leap' hasn't landed as promised, and autonomous AI research is undemonstrated. But the core compute and capability trends are real and on track, so it's not pure hype.

Real: compute scaling (~0.5 OOM/yr), trillion-dollar capex, ~83% on knowledge-work benchmarks. Overstated: a durable proprietary moat (open source didn't fade), near-term revenue, and a demonstrated path to fully autonomous AI research.