📊 Full opportunity report: Engineering Is Automated. Research Is the Residual. on ThorstenMeyerAI.com — validation score, market gap, and execution plan.

Recent developments in AI capabilities suggest that AI systems can now automate most core engineering tasks involved in AI research, leaving research as the residual challenge. This shift could accelerate AI development timelines significantly, according to recent analyses of benchmark data.

Multiple independent benchmarks—CORE-Bench, MLE-Bench, and kernel design advancements—show AI systems are nearing saturation in automating core engineering skills essential to AI research. For example, CORE-Bench, which measures the ability to reproduce research experiments, reached 95.5% completion by December 2025, with one author declaring it ‘solved.’ Similarly, MLE-Bench, assessing performance on Kaggle competitions, hit 64.4% in February 2026, approaching mid-tier human performance. These trends indicate that the bottleneck for AI research is shifting away from engineering tasks, which are increasingly automated, toward the more creative and conceptual aspects of research. While some aspects of research remain less automatable, the structural pattern suggests that automation may extend into research activities sooner than previously expected, potentially closing the residual gap faster than Clark’s initial thesis predicted.

Implications of Engineering Automation for AI Development Speed

The automation of core engineering tasks means that the primary remaining challenge in AI progress may soon be research itself. This could lead to faster development cycles, reduced costs, and a potential shift in how AI research is conducted. The findings challenge the notion that inspiration and creativity are the main barriers, implying that institutional and technological factors may become the new bottlenecks. If research can be automated, the timeline for breakthroughs could shorten dramatically, impacting industries, regulation, and global AI strategy.

Recent Benchmark Progress and AI Capabilities Milestones

Historically, AI research has been hampered by manual, labor-intensive engineering tasks such as reproducing experiments, optimizing kernels, and managing dependencies. Recent advances, documented by Thorsten Meyer and others, show that AI systems now handle these tasks with high reliability. The CORE-Bench, which measures research reproduction, improved from 21.5% in September 2024 to 95.5% in December 2025, with the ‘solved’ declaration marking a key milestone. Similarly, the MLE-Bench, testing on Kaggle competitions, rose from 16.9% in October 2024 to 64.4% in February 2026. Kernel design improvements, including automated GPU kernel generation, have become increasingly common, indicating a move toward production-grade automation. These developments suggest that the engineering component of AI research is approaching complete automation, shifting the focus toward the remaining research challenges.

“AI can today automate vast swatches, perhaps the entirety, of AI engineering. It is not yet clear how much of AI research it can automate, given that some aspects of research may be distinct from the engineering skills.”

— Thorsten Meyer

Remaining Uncertainties About Research Automation

While engineering tasks are nearing full automation, it remains uncertain how much of the research process—such as hypothesis generation, novel experimentation, and creative insight—can be automated. The structural question posed by Clark suggests that research may itself be a form of large-scale engineering, which could accelerate automation timelines. However, the degree to which current AI systems can handle the conceptual and innovative aspects of research is still under investigation. It is not yet clear whether the residual gap will close within the next 32 months or if unforeseen barriers will emerge.

Next Milestones in AI Research Automation

In the coming months, focus will likely be on extending automation into more abstract research activities, such as hypothesis formulation and experimental design. Monitoring the progression of advanced benchmarks and new research papers demonstrating kernel and algorithm innovations will be key indicators. Additionally, industry and academic institutions may begin to experiment with fully automated research pipelines, testing the limits of current AI capabilities. The pace of progress suggests that substantial automation of research could occur within the next two to three years, potentially transforming AI development practices.

Key Questions

What does automation of AI engineering mean for human researchers?

It suggests that many manual, repetitive tasks in AI research, such as reproducing experiments and optimizing code, can be handled by AI systems, allowing human researchers to focus more on creative and conceptual aspects.

Are all aspects of AI research automatable?

It is not yet clear if all research activities, especially hypothesis generation and innovative experimentation, can be fully automated. Current evidence indicates significant progress, but some elements may remain human-driven for now.

How soon could research automation impact AI development timelines?

If current trends continue, substantial automation of research activities could occur within the next two to three years, potentially accelerating breakthroughs and reducing development cycles.

What are the risks of automating AI research?

Potential risks include over-reliance on automated systems, reduced human oversight, and challenges in ensuring the quality and safety of automated research outputs. Careful regulation and oversight will be important.

Source: ThorstenMeyerAI.com