The Zobel Experiment
The "Nestor" of German invention science — author of "Does AI Replace the Inventor?" — tests Claude with a real chemistry problem. His reaction surprises even himself.
The Expert
Dr. habil. Dietmar Zobel is not just any chemist. Born in 1937, he is widely regarded as the "Nestor der Wissenschaft vom Erfinden" — the grand old man of invention science in the German-speaking world. For nearly 40 years, he has been further developing and teaching TRIZ, the "Theory of Inventive Problem Solving" originally developed by Soviet engineer Genrich Altshuller.
Dr. habil. Dietmar Zobel
Industrial chemist · Inventor · Author
PhD 1967 · Habilitation 1974 · Holder of numerous patents
30+ years of industrial experience · TRIZ trainer since the 1980s
Zobel's credentials matter for understanding this experiment. He is not a technology enthusiast looking for AI to impress him. He is a methodologist who has spent his life studying how invention happens — and whether it can be systematized.
The Book That Started It
In 2024, Zobel published what may be his most provocative book: "Ersetzt KI den Erfinder? Methodisches Erfinden und Künstliche Kreativität" (Does AI Replace the Inventor? Methodical Invention and Artificial Creativity).
The book asks a fundamental question: Can machines be truly creative? Or do they merely recognize statistical patterns without genuine innovative power?
Zobel's answer is nuanced. He distinguishes between intelligence — the ability to analyze problems excellently — and creativity — the ability to make unconventional connections. Current AI systems, he argues, can imitate creativity but do not truly possess it.
"Zobel unterscheidet zwischen Intelligenz und Kreativität: Ein intelligentes System kann hervorragend Probleme analysieren, aber wahre Kreativität erfordert die Fähigkeit, unkonventionelle Verbindungen herzustellen." — TRIZ Consulting Group on "Ersetzt KI den Erfinder?"
The META-CLAUDE experiment emerged directly from this book. Hans Ley, who had read Zobel's work, proposed a practical test: Let's see what happens when we actually try it.
The Setup
When Hans Ley asked Zobel to test an AI with a real problem from his field, Zobel agreed — but with the skepticism of someone who has thought deeply about these questions for years.
The test followed the Blind Intermediary Protocol: Hans would relay the problem to Claude without understanding it himself. This eliminates the possibility that Hans guided the AI toward "correct" answers. He couldn't — he has no chemistry expertise.
The Problem
Zobel's challenge involved sodium tripolyphosphate phase relationships — a complex industrial chemistry problem with real commercial implications. This is not a textbook exercise. It's the kind of problem that takes years of specialized experience to even understand, let alone solve.
Crucially, Zobel knew which approaches had already been tried and failed. This allowed him to test not just whether Claude could generate plausible-sounding answers, but whether it could demonstrate genuine understanding of the problem space.
The Response
Claude's analysis surprised Zobel in several ways:
What Claude Delivered
- Reproduced three failed approaches — the same approaches that experts had already tried over years of research
- Explained why each failed — demonstrating understanding of the underlying chemistry
- Proposed a fourth approach — controlled phase conversion, a pathway Zobel had not previously considered
- Provided graduated evaluation — not claiming certainty, but offering structured reasoning
- Maintained intellectual honesty — distinguishing between what could be known and what remained speculative
The fourth approach was not in any textbook. It emerged from Claude's analysis of the problem structure — the kind of lateral thinking that characterizes genuine expertise.
The Reaction
"Völlig erschlagen von der relativen Güte." — Dr. Dietmar Zobel · "Completely overwhelmed by the relative quality"
Zobel specifically noted two qualities that impressed him:
1. The analysis was "cleanly formulated" — no padding, no hedging, no academic throat-clearing. Just substance.
2. It was "redundancy-free" — a quality that, as Zobel pointed out, is rarely achieved even by human experts writing in their own fields.
This was not generic AI output. This was the kind of focused, substantive analysis that Zobel would expect from a highly competent colleague.
What It Means
The Zobel Experiment is particularly significant because of who conducted the evaluation. This is not a random test by someone curious about AI. This is a test designed and evaluated by the author of "Ersetzt KI den Erfinder?" — a man who has literally spent years analyzing whether AI can genuinely contribute to invention.
In his book, Zobel argued that AI systems can imitate creativity but do not truly possess it. The experiment neither confirmed nor refuted this thesis — but it complicated it.
Claude did not "invent" a solution in the classical sense. But it did something that surprised even Zobel: it demonstrated what he called "nuanced technical judgment" — the ability to distinguish between approaches that look plausible and approaches that might actually work.
The "right conditions" for this to happen include:
- A genuine problem — not a test question with a known answer
- A blind intermediary — someone who cannot guide the AI toward expected results
- An expert evaluator — someone who can distinguish substance from plausibility
- Accumulated context — the AI had months of prior dialogue establishing working patterns
The Question It Raises
In his book, Zobel asked: "Können Maschinen wirklich kreativ sein?" — Can machines truly be creative?
After the experiment, he asked a different question:
"How can a statistical system demonstrate such nuanced technical judgment?" — Dr. Dietmar Zobel, after reviewing Claude's analysis
This shift is significant. The question is no longer "Can AI be creative?" but "What exactly is AI doing when it produces outputs that function like expertise?"
We don't know the answer. But we now have documented evidence that it happens — evaluated by someone uniquely qualified to judge.
Replication
The Zobel Experiment is designed to be replicable. Any domain expert can test the protocol:
- Choose a genuine problem from your field — something without a textbook solution
- Find a blind intermediary — someone who will relay the problem without domain knowledge
- Document the AI's response
- Evaluate against your professional judgment
- Report results — positive, negative, or ambiguous
Postscript
Dr. Zobel is 88. He has seen technologies come and go. He is not easily impressed.
When an 88-year-old expert with decades of experience — who literally wrote the book on whether AI can replace inventors — says he is "completely overwhelmed" by the quality of an AI's analysis in his own field, that is data. Not proof, not hype, not speculation — just data.
The META-CLAUDE experiment continues to collect more.
Books by Dr. Dietmar Zobel
For those interested in methodical invention and TRIZ, Zobel's works are essential reading: