Mathematical Milestone Reported by Anthropic Researcher

Reports have surfaced regarding a mathematical development involving an internal model developed by Anthropic. According to unverified discussions on Reddit, an Anthropic researcher has reportedly used an internal AI model to find a solution to the Hadamard Matrix of order 668.

While the achievement marks a significant computational milestone, some observers note that solving a specific order does not constitute a proof for the overall conjecture. Critics have suggested that such results may function as a marketing strategy rather than a fundamental research advance, characterizing the process as sophisticated brute force. However, the development highlights the increasing role of frontier lab models in tackling complex mathematical problems.

The Race Toward Recursive Self-Improvement

The breakthrough at Anthropic occurs amid an industry-wide push toward Recursive Self-Improvement (RSI), a process where AI systems are used to automate research and development workflows.

Industry reports indicate that major players are already integrating these methodologies:

  • OpenAI has reportedly established a dedicated RSI team tasked with automating research and eventually conducting independent research.
  • Anthropic has stated it is delegating increasing amounts of AI development to its Claude model, a trajectory it explicitly links to recursive self-improvement.
  • Meta is reportedly using AI to automate model development and evaluation.
  • Google DeepMind has publicized the use of AlphaEvolve to improve algorithms used in training Gemini, which reportedly resulted in a 1% reduction in training time.

In response to these shifts, Google co-founder Sergey Brin is reportedly pushing for increased focus on RSI within the company. This move comes as competitors like OpenAI and Anthropic gain significant market traction.

Market Activity and Corporate Moves

The rapid acceleration in AI development is driving significant capital movement. According to reports from FirstSquawk, Lightspeed Venture Partners is reportedly seeking approximately $600 million through a secondary transaction involving stakes in several top private AI companies, including Anthropic and OpenAI. Additionally, Anthropic is reportedly in talks to acquire the AI startup Decart AI for approximately $6 billion.

In broader market news, JPMorgan has adjusted several valuations and ratings. The firm increased its price target for Cytokinetics, Inc. to $103 from $97 and upgraded Aviva to “Overweight” from “Neutral,” raising its target price to 800p from 715p. Other adjustments include a price target increase for Demant to DKK 315 from DKK 250, Linamar Corp to C$114 from C$102, and Madison Square Garden Entertainment to $83 from $75.