Particle.news
Get it on Google Play
Download on the App Store
32 ARTICLES
·
1h ago

Anthropic Says Self‑Improving AI May Be Imminent as Claude Now Writes Most Code

The disclosure signals a widening gap between accelerating AI self‑development and the industry's readiness to manage safety, governance, infrastructure.

The Nasdaq MarketSite in New York on October 16, 2026.
Stickers for Claude software on a laptop at a conference in London on May 19, 2026.
Anthropic CEO Dario Amodei.
Image

Overview

  • Anthropic told the SEC in an amended S-1 filed on Nov. 2 that internal estimates indicate it may "imminently" meet its Responsible Scaling Policy threshold for "dramatically accelerated progress attributable to AI," which is its definition of recursive self‑improvement.
  • The filing clarifies that no current model is designated as having crossed the threshold and that the company has failed to spot past accelerations until after they occurred.
  • Anthropic reported that its Claude systems produced over 90% of internal code contributions in September 2026, showing a rapid shift to AI‑driven engineering work.
  • The company disclosed roughly $895.4 billion in aggregate compute and infrastructure commitments, a spike in compute costs that pushed Q3 compute expenses sharply higher even as reported revenue reached about $17 billion for the quarter.
  • Institutional investors are pressing for more conservative IPO pricing and are closely watching Anthropic's governance and safety commitments because founders and a Long‑Term Benefit Trust retain controlling voting power.