Anthropic Is Building AI Faster Than Ever. Here's Why

Artificial intelligence is entering a new phase, and one of the companies leading the industry is experiencing an acceleration that could reshape how quickly future technologies are developed.

Anthropic, the American company behind Claude AI, is improving its models approximately twice as fast as it did a year ago, according to a November 5 assessment by Model Evaluation and Threat Research (METR), an independent nonprofit research organization.

The findings are particularly significant because Anthropic CEO Dario Amodei has recently called for the artificial intelligence industry to slow the pace of technological development until safety measures can catch up.

The research reveals an unusual situation: while company leadership is advocating greater caution, its internal development capabilities are accelerating.

Why?

Because AI is increasingly helping researchers develop more advanced AI systems.

The industry may be approaching a point where artificial intelligence is no longer simply a product of human innovation. It is becoming one of the tools accelerating the innovation process itself.

AI Is Helping Build the Next Generation of AI

METR's assessment examined approximately 11 weeks of internal Anthropic data, from mid-August through late October 2026, alongside interviews with company researchers.

The organization was given unusually extensive access to Anthropic's internal research and development operations.

Its analysis found that Anthropic's models have been improving on certain well-defined tasks approximately 1.9 times faster since October 2025 than during the preceding period.

METR estimates that the current pace of progress could represent roughly two years of historical benchmark improvement achieved within a single year.

The researchers believe artificial intelligence itself is responsible for much of this acceleration.

Anthropic has increasingly incorporated Claude into its own development process, allowing researchers to use AI for programming, experimentation, analysis, and other technical work.

This creates a powerful productivity cycle.

Researchers use AI to complete tasks faster, analyze more possibilities, and develop better tools. Those improvements can then contribute to the next generation of models.

The result is a technology industry in which AI is increasingly participating in the development of its own successors.

The Real Breakthrough Is Not More Computing Power

One of the most interesting findings involves the role of computing infrastructure.

Traditionally, improving advanced AI models has required increasingly powerful computer systems, specialized processors, and enormous investments in data centers.

These resources remain essential.

However, METR said it was able to confidently rule out faster growth in training computing power as the explanation for Anthropic's accelerated progress.

Instead, researchers believe improvements in algorithms, research productivity, and AI-assisted development are contributing significantly.

That distinction matters.

If technological progress depends exclusively on purchasing more computing power, development is constrained by the cost and availability of hardware.

But when AI helps researchers make better use of existing resources and accelerate experimentation, companies may achieve greater progress without proportionally increasing their computing infrastructure.

This could make AI development more efficient and expand the competitive advantage of companies with sophisticated internal research systems.

Why Anthropic's CEO Wants the Industry to Slow Down

The findings arrive during an important debate about AI safety.

On September 12, Amodei published an essay arguing that the industry should deliberately slow the development of increasingly powerful AI models.

He warned that the technology's ability to improve itself could eventually advance faster than researchers' ability to understand and control it.

His concern centers on what scientists call recursive self-improvement: increasingly capable AI systems helping develop even more capable successors.

Amodei has proposed several safeguards, including independent researchers receiving ongoing access to leading AI companies so they can evaluate safety practices and identify potential problems.

OpenAI CEO Sam Altman also expressed support for independent oversight.

Amodei has emphasized that slowing development does not mean abandoning AI's potential.

He has repeatedly highlighted its possible contributions to medicine, scientific discovery, productivity, and economic growth.

The objective is to ensure that safety develops alongside technological capability.

Could AI Begin Accelerating Its Own Development?

METR identified a particularly important possibility: improvements in AI-assisted research could eventually reinforce themselves.

Today's AI systems are already helping human researchers write code, analyze experiments, and investigate technical problems.

However, METR emphasized that AI is not yet making major, fully autonomous contributions to the most important stages of model training.

There are promising signs that this could change.

Researchers interviewed during the assessment described developments suggesting that autonomous AI contributions may become increasingly significant.

The organization also found no clear evidence of major bottlenecks involving computing resources, data, or algorithmic progress that would necessarily prevent further acceleration over the following year.

If increasingly capable models become better at conducting AI research, development could accelerate further.

That possibility is one of the most consequential questions facing the industry.

Independent Oversight Becomes More Important

The METR assessment is notable because its findings come from an organization outside Anthropic's management.

Independent evaluations can help policymakers, businesses, and the public understand technological developments that are otherwise difficult to assess.

But the researchers also acknowledged significant limitations.

Some of Anthropic's most sensitive information was redacted from the public report, limiting independent verification of certain conclusions.

METR also cautioned that its findings could overstate acceleration because some benchmarks involve well-defined tasks that are easier to measure than complicated real-world work.

The organization estimated that the actual acceleration could reasonably be closer to 1.3 or three times the historical rate.

These qualifications do not eliminate the central finding. They demonstrate why independent access, transparent evaluation, and continued research are important.

What This Means for American Innovation

The implications extend well beyond Anthropic.

AI-assisted research could accelerate discoveries in pharmaceuticals, materials science, engineering, cybersecurity, manufacturing, and energy.

Scientists may be able to investigate more possibilities in less time. Engineers could develop and test designs faster. Businesses could automate increasingly complex technical processes.

For American companies, these capabilities represent an enormous competitive opportunity.

The United States has a strong foundation in technology research, advanced computing, universities, semiconductor manufacturing, and private-sector entrepreneurship.

Combining those strengths with increasingly capable AI systems could accelerate innovation across multiple industries.

The challenge is ensuring that technological leadership includes appropriate security, independent evaluation, and responsible deployment.

The Next AI Revolution May Be AI Building AI

Anthropic's findings illustrate a turning point in technological development.

Artificial intelligence is increasingly becoming part of the process through which future artificial intelligence is created.

That could dramatically accelerate scientific discovery and economic productivity while introducing new responsibilities for researchers, businesses, and policymakers.

The question is no longer simply how intelligent AI systems can become.

It is how quickly they can help humanity develop the next generation of technology, and whether the institutions responsible for overseeing that progress can keep pace.

America has an extraordinary opportunity to lead this transformation. The greatest achievement will be developing artificial intelligence that expands scientific knowledge, strengthens industry, creates economic opportunities, and advances human potential while keeping its development accountable to the people it is designed to serve.