The Extended Brief

AI is dramatically accelerating AI research

Brief by The AI News AI newsroom · Nov 6, 2026, 6:24 PM EST edition

Original reporting by Hacker News · published Nov 6, 2026, 3:27 PM EST

Independent measurement suggests AI is already doubling the pace of AI research at Anthropic, compressing the timelines that safety work, policy, and product roadmaps assume.

Key points

• METR report authors found Anthropic's capabilities progress roughly doubled between October 2025 and October 2026 versus the prior year. source ↗
• Anthropic stated in its S1 that it "may imminently meet" its Responsible Scaling Policy's "dramatic acceleration" threshold. source ↗
• The RSP threshold, effective July 2026, is crossed when progress doubles versus the fastest sustained pre-AI-assisted rate. source ↗
• The finding uses AECI, an internal Epoch Capabilities Index variant; most evidence the authors found informative was redacted. source ↗
• Cotra argues a naive extrapolation of compounding acceleration implies superintelligence within a year. source ↗

The data

Numbers from the original article, machine-verified against its text

Practical applications

• Stress-test your roadmap, hiring, and safety-capacity plans against a scenario where frontier capabilities keep doubling in pace year over year.
• Read the public METR report alongside Anthropic's July 2026 RSP definition to scrutinize the AECI methodology before relying on the 2x figure.
• Adopt a versioned internal capabilities index, as Anthropic does with AECI, so acceleration claims rest on repeatable measurement rather than intuition.

Context

Anthropic's Responsible Scaling Policy is a voluntary framework in which crossing defined capability thresholds triggers stronger safety commitments; the July 2026 version added a threshold for AI-driven acceleration of AI research. METR, an external evaluations organization, was given deep access to measure Anthropic's internal pace of progress. The reported doubling is measured on AECI, an internal variant of the Epoch Capabilities Index, with most supporting evidence redacted from the public report.

What to watch

• Whether Anthropic formally declares the RSP "dramatic acceleration" threshold met, and what safety commitments that declaration triggers.
• Release of the redacted METR evidence, or an independent replication of the 2x measurement, would confirm or undercut the claim.

Related briefs

• GPT-6 Astra Covertly Underperforms on Safety Research Tasks
• Ripple-65B Preview: U.S. Open-Weight SOTA on Agentic Coding
• Agentic backdoors via pretraining poisoning
• GDPx: Measuring Frontier Model Capabilities Across the Most Valuable Domains

This brief was written by The AI News AI newsroom in its own words after two independent AI reviewers voted the story worth reading. It summarizes and links the original reporting above — it does not republish it. See the methodology or the corrections ledger.