In parallel with this interface overhaul, the company has published a series of metrics intended to monitor progress in developing AI systems at advanced laboratories. These measures focus on three essential pillars: the volume of research and development carried out exclusively by AI systems, the quality of supervision exercised over autonomous agents, and the methods of allocating computational resources. The observed trend indicates an exponential increase in the power of these systems, which have begun to automate their own construction and refinement processes. According to the presented data, the Claude model is now responsible for 26% of the company’s research and development efforts. This figure represents a spectacular increase from February 2026, when its contribution was considerably smaller. Analysis of the trajectory suggests that, if the current pace continues, Claude could lead 100% of research and development processes by February 2028.
Some commentators argue that this acceleration speed surpasses even the most optimistic prior predictions regarding Recursive Self-Improvement (RSI). Transparency Shows AI Building Its Own Successors Another notable aspect is that, for the first time, Anthropic has offered detailed transparency into the speed at which AI is building itself. The data show that the proportion of work performed by AI in constructing the next generation of models rose from below 1% to over 26% in just a few months. Internal projections suggest this percentage could reach 80% by the end of the current year. Currently, approximately 30,000 AI agents operate simultaneously within the company’s infrastructure to support these processes.
Although the data are encouraging, some experts remain skeptical about the clear definition of the scope of „AI research and development.” Critics argue that the lack of a precise delineation of the tasks included in the statistic prevents a complete evaluation of the RSI phenomenon. Nevertheless, many acknowledge that this step toward transparency is significant progress in the right direction. Debate Emerges Over Pacing and Control in AI Development Another point of discussion centers on the concept of „Pacing” in AI development. Some analysts claim that Anthropic is adding an ad-hoc contribution to this issue, emphasizing that the pacing process is already underway, albeit loosely structured and disordered.
It is suggested that a more rigorous approach is needed, comparable to the discipline required in quantitative trading, to manage technological acceleration more effectively. Practical Scenario Shows Claude Managing Data Pipeline Failures The practical scenario described by the company illustrates how, in the case of a data pipeline failure, a human can delegate the task to Claude. The system checks the logs, identifies the cause of the problem, writes remediation code, runs the necessary tests, and continues operations without interruption, even when new issues arise. OpenAI Staff Expect Hodge Conjecture Solution Soon After Navier-Stokes Breakthrough Beyond technical updates, another piece of news has been made public from the mathematics and AI industry. OpenAI employees expect the Millennium Problem known as the Hodge Conjecture to be solved relatively soon, following the already resolved Navier-Stokes equations.
This event recalls tensions from earlier this month, when OpenAI was criticized by mathematicians for attempting to claim credit for solving an extremely difficult problem, sparking heated debates in the academic community. The company has confirmed that it has made significant progress in solving a second million-dollar mathematical problem. Some specialists describe the company’s behavior as „predatory,” suggesting a conflict exists between mathematical elites and OpenAI’s aggressive approach to claiming solutions to prize-winning problems. Huawei Chief Urges Chinese AI Researchers to Accelerate to Identify Existential Risks In a contrasting note, Eric Xu, chairman of Huawei, declared that Chinese AI researchers must accelerate development speed to identify the existential risks of technology. This position stands in direct opposition to calls from Silicon Valley for a prudent slowdown in development, in a context where concerns about existential risks are constantly growing.
The executive’s comments were widely discussed in the press, including in publications such as Reuters, the Washington Post, and the Financial Times, highlighting the strategic divergence between the two technological superpowers in the face of challenges posed by artificial intelligence.