Watching the plot · updated September 7, 2026

Is AI helping yet?

The metric for AI is whether it's improving the lives of conscious creatures. AI's capability is going vertical. The things that actually matter for human beings are mostly still improving — slowly, on their own long curves. Here are both, for the whole planet, side by side.

The AI curve

Frontier training compute has climbed by roughly nine orders of magnitude since 2012 — a billionfold. What a model can do in one sitting is doubling about every seven months. This column is a set of hockey sticks.

The human curve

Extreme poverty, child mortality, and hunger keep falling; life expectancy, literacy, and electricity keep rising. Real progress — but gentle slopes, and no visible bend where the AI curve took off.

AI capability

what machines can now do — Epoch AI, METR

Frontier model training compute

log₁₀ FLOP (largest model that year)
10^26.7 2025 +9 orders since 2012

The largest training compute (in floating-point operations) of any notable model published that year, on a log scale. AI capability is a story of compute.

2012–2025 · Epoch AI — Notable AI Models database

Notable AI models released

models per year
107 2025 ▲ 723% since 2012

Count of models Epoch AI classifies as notable (state-of-the-art, historically significant, or highly cited) with a publication date in that year.

2012–2025 · Epoch AI — Notable AI Models database

AI task time-horizon (50% reliability)

human-minutes of task length
130 2025 ×4,300 since 2019

The length of task (in the time it takes a human) that a frontier model can finish with 50% reliability. Doubling roughly every 7 months. Frozen citation read from METR's published figure.

2019–2025 · METR — Measuring AI Ability to Complete Long Tasks

Human progress

whether life is getting better — World Bank, World Happiness Report

Child mortality (under 5)

deaths per 1,000 live births
37.4 2024 ▼ 60% since 1990

1990–2024 · World Bank / UN IGME

Undernourishment

% of world population
8.50 2023 ▼ 34% since 2001

2001–2023 · World Bank / FAO

What we can't measure yet

The most important part of the question is the hardest to source. These have no clean, redistributable global time series — so the honest move is to name them, not fake them. This is the frontier of the site.