Where are we now

April 2026

AITAI's assessment of current AI trajectory

Capability is advancing rapidly across benchmarks, coding, and reasoning. But adoption remains shallow and uneven — concentrated in large firms, specific sectors, and a handful of countries. Australia's public investment lags significantly.

The current trajectory points toward:

Capability: Mid, leaning towards fast
Diffusion: Slow and scattered

Supporting evidence

Key signals across both axes

Capability

48.9pp jump on GPQA in one year

Stanford HAI, 2026

SWE-bench: single digits to 70%+

SWE-bench, 2025

Training compute doubling every 6 months

Epoch AI

Inference cost dropped 10x in 24 months

Industry pricing data

Open-weight gap closed from 24 months to 6–12

Stanford HAI, 2026

Diffusion

11% direct US workplace adoption — stalled

Hartley et al., SSRN 2026

OECD firm deployment at 20.2%

OECD, 2025

Only 4% of occupations deeply integrated

Anthropic Economic Index, 2026

Gen AI: 53% adoption in 3 years — fastest GPT ever

Stanford HAI, 2026

Australia: early and uneven

JSA, 2024

A$0.3B government investment vs A$13.8B comparator average

ATSE, 2024

Policy robustness analysis

Which of WA's five key policy levers hold up across all six scenarios?