Capabilities · Trend
Reasoning models
Models trained to deliberate before answering turned test-time thinking into a new scaling axis.
Forecast horizon
Deliberate reasoning becomes default; the question shifts to cost, faithfulness and verifiable chains.
Connections
Connections · 5
How this node ties into the rest of the map, and the evidence behind each link.
from · requires 8/10
AI agents Agents lean on deliberate reasoning to plan multi-step actions.
+3 growth to · uses 8/10
Test-time compute Reasoning models convert inference compute into accuracy.
+2 growth from · enables 7/10
Compute scaling More compute underwrites the reasoning paradigm.
+2 growth to · enables 6/10
AI for science Stronger reasoning unlocks math and scientific problem-solving.
+2 growth from · enables 6/10
Synthetic data Verifier-generated data trains better reasoning.
+2 growthSignal 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.
METR → 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.
Epoch AI →