> ## Content Index
> Fetch the complete content index at: https://adjacent.media/llms.txt
> Use this file to discover other available public pages before exploring further.

# AI labs are now using AI to accelerate their own research
- URL: https://adjacent.media/signals/ai-labs-are-now-using-ai-to-accelerate-their-own-research/
- Published: 2026-11-11T00:17:18.000Z
- Updated: 2026-11-11T00:17:18.000Z
- Description: Anthropic has doubled its model improvement rate on defined tasks since early 2024, with METR’s analysis suggesting AI systems are now materially speeding up the company’s own R&D cycle—crossing a threshold the company itself flagged as “dramatically accelerated progress.
- Author: Jonathan Greene
- Tags: #signal, theme-ai, ai capability acceleration, recursive self-improvement, ai safety

Source: [Hyperdimensional](https://open.substack.com/pub/hyperdimensional/p/iron-makes-iron)

Anthropic has doubled its model improvement rate on defined tasks since early 2024, with METR's analysis suggesting AI systems are now materially speeding up the company's own R&D cycle—crossing a threshold the company itself flagged as "dramatically accelerated progress." Labs that can deploy their best models as research tools gain compounding advantages in capability gains, making it harder for slower-moving competitors to catch up. This raises real questions about whether external oversight can keep pace with acceleration that's now self-reinforcing. The safety community's visceral reaction reflects genuine uncertainty about whether existing safeguards like Anthropic's Responsible Scaling Policy remain meaningful when progress itself becomes difficult to measure and the incentive to deploy AI-as-researcher conflicts with transparency.