The Falsification Gap: Why Slow-to-Test Work Is Easy to Fake
Updated
Knowledge on this page was mainly distilled from Product Managers Are About to Be Found Out.
The falsification gap is the delay between when work is produced and when reality can prove it wrong. The wider the gap, the easier AI can imitate the work convincingly.
Code has a narrow falsification gap. A generated function that looks right gets caught by the test suite, the build, or the 2 a.m. incident. The feedback loop is short and indifferent to confidence.
Product strategy, roadmaps, prioritization frameworks, and stakeholder memos have a wide falsification gap. A roadmap can be wrong and still sound mature. A strategy memo can be completely generic and still get nodded through a room of smart people. The verdict arrives a quarter later, quietly, after the document is forgotten and its author has written three more.
AI thrives inside this gap. It produces the convincing surface faster, cheaper, and without end. The same quality that let weak work survive in ambiguous domains is the quality that makes it trivial to automate.
Q&A
What is the falsification gap?
It is the delay between when work is produced and when reality can prove it wrong. Code has a narrow gap because tests, builds, and production failures surface errors quickly. Strategic and analytical work has a wide gap because the verdict may not arrive for months. The wider the gap, the easier it is for AI to produce output that passes for competent human work.
Why does a wide falsification gap make work vulnerable to AI?
AI generates convincing surfaces by pattern-matching on large volumes of existing work. When reality checks are slow or absent, there is no mechanism to distinguish AI-generated output from thoughtful human judgment. The work only needs to pass the taste of the people in the room, not a concrete test against reality.
How do strong practitioners close the falsification gap deliberately?
They manufacture fast, concrete tests. A fake door that counts who clicks through. The thinnest possible slice shipped to ten real users. A pre-order taken before a line of code exists. Each one drags a verdict from next quarter into this week. These self-imposed tests build the feedback loop that ambiguous work never had.
Does the falsification gap apply only to product management?
No. It applies to any knowledge work where output is judged by how it reads rather than what it does. Consulting decks, marketing strategies, policy memos, and research summaries all share wide falsification gaps. The framework predicts that roles producing those outputs face higher AI substitution risk than roles where output is tested quickly and concretely.
Is the falsification gap the same as ambiguity?
Not exactly. Ambiguity is the medium in which certain work happens. The falsification gap is the time it takes for reality to resolve that ambiguity. Good practitioners resolve ambiguity quickly by seeking contact with reality. Weak practitioners hide inside it, producing work that sounds correct but is never concretely tested.