📊 Full opportunity report: Engineering Is Automated. Research Is the Residual. on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
AI systems have achieved near-saturation in core engineering skills, automating most engineering tasks. However, research activities are still predominantly performed by humans, though this may change soon.
Recent AI capability benchmarks demonstrate that AI systems can now automate the majority of core engineering tasks, with some research activities remaining human-driven. This marks a significant shift in AI’s role in R&D, potentially transforming the landscape of technological development.
According to Thorsten Meyer’s analysis of Jack Clark’s recent essay, six key benchmarks measuring AI’s progress in core science and engineering skills are approaching or have achieved saturation. For example, the CORE-Bench, which tests AI’s ability to reproduce research, has risen from 21.5% in September 2024 to 95.5% in December 2025, with the benchmark’s author declaring it ‘solved.’ Similarly, the MLE-Bench, evaluating AI performance on Kaggle competitions, has advanced from 16.9% in October 2024 to 64.4% in February 2026, with AI now reaching mid-tier human performance levels. These trajectories suggest that the engineering side of AI R&D—such as reproducing experiments, optimizing kernels, and designing infrastructure—is nearing full automation.
In contrast, Clark and Meyer note that research activities—such as formulating hypotheses, creative problem solving, and theoretical innovation—are less amenable to automation at this stage. While some progress is evident, the consensus is that research remains largely a human endeavor, though the structural question remains open whether research itself is becoming a form of engineering at scale. The critical implication: the bottleneck on innovation may shift from engineering to research, but this transition is not yet complete.
Engineering is automated.
Research is the residual.
Six skill benchmarks. Edison’s framing. The question Clark leaves open is whether research is just engineering at scale.
Jack Clark’s Import AI #455 catalogs six benchmarks measuring AI capability on AI R&D tasks and concludes “AI can today automate vast swatches, perhaps the entirety, of AI engineering.” The residual question is research. The structural read on the residual: it may not be a permanent moat.
Six skills. One trajectory.
Clark catalogs six benchmarks measuring AI capability on AI R&D-relevant tasks. Each individual benchmark could be noise. Six benchmarks moving together is a curve. The pattern is the cascade observed across the broader Clark series — visible here in the specific R&D-skill domain.
CLAUDE AI UNLEASHED From First Prompts to Pro: The Complete Guide to Claude AI for Writing, Research, Coding, and Business (The Claude AI Mastery Series)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Three data points. Mixed signal.
Clark provides three data points on the creative-spark question. Yes-evidence: Erdős-1051, centaur math discovery, sporadic Move-37-style moments. No-evidence: low yield, framing dependence, absence of acceleration. The mixed signal is the honest read.
The data supports two readings. Pessimistic: rare moments suggest creative insight is qualitatively distinct from engineering work. Optimistic: rare moments are an artifact of low-volume exploration; more shots on goal yields more discoveries. Both readings are consistent with Clark’s “vast swatches, perhaps the entirety” claim. They differ on the residual.
Concurrent Engineering: Automation, Tools, and Techniques
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Five dimensions Clark gestures at but leaves underdeveloped.
Clark’s section is rigorous on the empirical evidence. Five strategic dimensions matter for the institutional response that the Clark series synthesis argues is structurally inadequate.
AI research hypothesis generator
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Two readings. Different equilibria.
The structural question Clark leaves open: is research a permanent moat that bounds automated AI R&D, or is it engineering at scale that dissolves with more shots on goal? Both readings are consistent with the current data. They differ by orders of magnitude in consequences.
Productivity multiplier years
Recursive loop operational
The AI-Powered Scholar: Transform Every Stage of Your Research With Cutting-edge AI (AI-Powered Research Toolkit — A Mastering Research Series)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Five audiences. Asymmetric cost of being wrong.
The institutional response should not bet on inspiration being a permanent moat. If the distinction holds, capacity built is still useful. If it closes, capacity is necessary. Asymmetric cost-of-being-wrong points toward building now.
IN INDUSTRY
IN ACADEMIA
POLICYMAKERS
INVESTORS
EVERYONE ELSE
Engineering is automated. The residual is the question. The institutional response should not bet on inspiration being a permanent moat.
Implications for AI-Driven Innovation
The near-complete automation of engineering tasks could dramatically accelerate technological development, reducing costs and timeframes for research and product deployment. If research activities follow the same trajectory, the pace of innovation could increase exponentially, reshaping industries and R&D processes. However, the current gap in automating research means that human creativity and hypothesis generation still dominate, possibly limiting AI’s impact in the short term. Understanding whether research becomes automatable at scale is crucial for predicting future AI capabilities and industry shifts.
Recent Advances in AI R&D Capabilities
Over the past 18 months, multiple benchmarks have shown rapid progress in AI’s technical skills relevant to R&D. The CORE-Bench, measuring research reproduction, has seen a 4.4× improvement, with AI systems now capable of handling dependencies, code execution, and output analysis at near-human levels. Similarly, the MLE-Bench, assessing performance on Kaggle competitions, has advanced from beginner to mid-tier human performance, with AI reaching about two-thirds of the human medalist level. These benchmarks are part of a broader pattern where AI models are increasingly capable of performing complex engineering tasks, including kernel design, infrastructure optimization, and code generation, often demonstrated through research papers and open-source projects.
Clark’s analysis suggests that these improvements are approaching the limits of current measurement tools, indicating that the underlying capabilities are reaching a saturation point. This pattern aligns with the broader hypothesis that AI’s engineering skills are becoming fully automated, while research remains a more complex, less-measurable domain.
“The structural read is that research may itself be engineering at scale — in which case the residual closes faster than Clark’s framing implies.”
— Thorsten Meyer
Unexplored Aspects of AI Research Automation
It remains unclear how soon AI will fully automate research activities such as hypothesis generation, theoretical innovation, and creative problem solving. The structural question Clark leaves open—whether research itself is a form of engineering—has not yet been definitively answered. Moreover, the pace at which research automation could occur depends on future breakthroughs and institutional adaptations, which are still uncertain.
Next Milestones in AI R&D Capabilities
In the coming 12 to 24 months, researchers will likely continue to push the boundaries of AI’s engineering skills, with benchmarks approaching saturation. Attention will turn to whether new metrics or approaches can measure progress in research automation. Industry and academia will also examine how to adapt workflows to leverage AI fully, potentially shifting R&D structures. Monitoring developments in AI-generated hypotheses, code, and experimental design will be critical for assessing the pace of research automation.
Key Questions
Will AI fully replace human researchers in the near future?
While AI is rapidly automating engineering tasks, research activities involving creativity, hypothesis formulation, and theoretical innovation are still largely human-driven. The timeline for full automation of research remains uncertain.
What are the risks of relying on AI for engineering and research?
Potential risks include over-reliance on AI-generated outputs, lack of transparency, and the possibility of missing novel insights that require human intuition. Ethical and safety considerations also remain important.
How might industries adapt to increased AI automation in engineering?
Industries may shift towards more AI-assisted workflows, reduce costs, and accelerate development cycles. They will also need to develop new skills and oversight mechanisms to manage AI-generated outputs.
What technological breakthroughs are needed to fully automate research?
Advances in AI creativity, hypothesis generation, and understanding of scientific principles are required. Developing metrics to measure research automation progress is also essential.
Source: ThorstenMeyerAI.com
