Andreas Dahlström’s Manifesto on Building Knowledge Machines in the AI Age (Thalius.ai, March 2026)

Andreas Dahlström, founder of Thalius.ai, authored a manifesto in March 2026 addressing founders, builders, and investors on navigating the AI-driven transformation in software development. He argues that human programming scarcity has collapsed due to AI’s ability to quickly generate code and reverse engineer features, rendering traditional software moats based on architecture obsolete. The deepest competitive moat now lies in sequential learning—a process that accumulates unique, hard-to-replicate knowledge over time—and in protecting proprietary learning loops rather than secret architectures. Dahlström emphasizes that knowledge machines, which continuously improve decision-making parameters based on private data and human feedback, are superior to static products. The manifesto introduces the 95/5 rule, where AI handles 95% of volume work and humans provide 5% critical judgment, which is essential for reliable knowledge work. The document stresses embracing open architectures and focusing defensibility on data-driven parameters and continuous learning loops. It applies to technology founders and investors preparing for an AI era before the advent of Artificial General Intelligence (AGI). This document excludes detailed AI technical algorithms or fully automated AGI scenarios.