METR: Anthropic's models are improving about twice as fast as they did between March 2024 and October 2025, with internal AI use driving ~80% of acceleration (Tiffany Li/The Information)
Recent findings reported by Tiffany Li for The Information show Anthropic's systems are advancing roughly twice as fast as their previous pace measured between March 2024 and October 2025. METR discovered that the organization's reliance on its own internal artificial intelligence accounts for roughly 80% of this rapid acceleration. This heavy dependence on in-house AI tools illustrates how foundational systems are directly speeding up the creation and refinement of subsequent models.
Key points
Deploying existing artificial intelligence tools internally serves as the primary driver…: Deploying existing artificial intelligence tools internally serves as the primary driver for faster capability gains.
The pace of machine learning advancement has doubled compared to the period spanning earl…: The pace of machine learning advancement has doubled compared to the period spanning early 2024 to late 2025.
Relying on automated systems internally plays a significant role in accelerating the crea…: Relying on automated systems internally plays a significant role in accelerating the creation of newer models.
Sources checked by our analysis pipeline: 1
Dina Bass / Bloomberg: |
Wall Street Journal: |
Emma Roth / The Verge: |
Andrew E. Freedman / Tom's Hardware: |
Jaron Schneider / PetaPixel: |
Tom Warren / The Verge: |
Takashi Mochizuki / Bloomberg: |
Andrew Webster / The Verge: |
Brody Ford / Bloomberg: |
New York Times: |
Tom Warren / The Verge: |
Takashi Mochizuki / Bloomberg: |
OpenAI: |
Isabelle Bousquette / Wall Street Journal: |
Tiffany Li / The Information: |
Carl Franzen / VentureBeat: |
Kelly Cloonan / Wall Street Journal: |
Radhika Rajkumar / ZDNET: |
Emma Roth / The Verge: |
Marcus Mendes / 9to5Mac: |
Ivan Mehta / TechCrunch: |
Ben Schoon / 9to5Google: |
Maxwell Zeff / Wall Street Journal: |
Rya Jetha / Business Insider: |
Jay Peters / The Verge: |
Reuters: |
Jonathan Stempel / Reuters: |
Charles Rollet / Business Insider: |
Los Angeles Times: |
David Keohane / Financial Times: |
Bloomberg: |
Mark Gurman / Bloomberg: |
This is a Techmeme archive page. It shows how the site appeared at 2:05…
Speaking at the WSJ Tech conference, corporate leadership acknowledged that the tech industry has struggled to communicate the benefits of its physical infrastructure to the public. As consumer apprehension surrounding automated systems grows, these large-scale facilities have become targets for political pushback and local community resistance. To protect ongoing multi-state investments, future project proposals will require strategic adjustments to secure bipartisan approval and address civic demands.
Advanced artificial intelligence tools are accelerating vulnerability discovery, enabling rapid detection of severe flaws across vital open-source environments like Linux and Firefox. While developers offer defensive initiatives and financial backing, high computing costs and restricted platform access threaten to tilt the advantage toward malicious actors. The emerging dynamic highlights an urgent race between resource-constrained defenders and novice attackers leveraging powerful models.
Meta is rolling out real-time conversational capabilities alongside synchronized animated avatars to enhance user immersion across its digital assistant ecosystem. This interaction model lets users customize vocal tones as the company prepares to transition its artificial intelligence software onto wearable hardware. The launch represents a critical step in expanding the reach of multimodal assistance ahead of new consumer device releases.