← Back to the trend map

Capabilities · Concept

Test-time compute

Spending more compute at inference buys accuracy — a second scaling law alongside training.

Trend strength 7/10

Momentum +2/q

Confidence high

Status rising

Forecast horizon

Adaptive compute budgets per query make 'how hard to think' an explicit governance dial.

Connections

Connections · 2

How this node ties into the rest of the map, and the evidence behind each link.

from · uses 8/10

Reasoning models

Reasoning models convert inference compute into accuracy.

from · uses 7/10

Agents spend inference compute to search and self-correct.

Signal sources

Signal sources

Dated facts from primary sources in this direction.

Task horizon doubling Mar 2025

The length of software tasks AI agents can do autonomously at 50% reliability has doubled about every 7 months — and since 2024 closer to every ~3 months.

Benchmarks saturating Apr 2025

In one year scores rose by 18.8, 48.9 and 67.3 points on MMMU, GPQA and SWE-bench; real-world software solve rate jumped from 4.4% to 71.7%.

Stanford HAI — AI Index 2025 →

Autonomous coding 2025–2026

On SWE-bench Verified (500 real GitHub issues), autonomous coding agents reached ~80–86% by late 2025, up from under 50% in early 2025.