Five Dated Predictions on Meaning, Machines, and Human Thinking (2027–2032) — with Public Annual Review (v1.0)

Five dated, falsifiable predictions on meaning, machines, and human thinking (2027–2032), version 1.0, bilingual English/German — with a standing commitment to public annual review every September, beginning September 2027. Each prediction carries an explicit deadline and an explicit failure condition: (1) major agent-interoperability efforts adopt concept-level meaning-versioning by end-2028; (2) a recognized standards body publishes a specification for versioned, resolvable concept identifiers for AI systems by end-2030; (3) person-specific fidelity testing becomes an established evaluation category with at least three independent benchmarks by end-2030; (4) at least five organizations publicly report a human-capability measure alongside AI-productivity measures by end-2029; (5) a leading world-model system exposes an explicit, human-readable semantic interface by end-2032. Review rules are fixed in the document: wording frozen, failed predictions remain failed, public evidence only, and the author's own projects do not count as evidence for predictions 2, 3 and 5. The predictions extend theses the author has published openly (Universal Concept Layer; the Ehstand test; the semantic layer for world models). Licence note: CC BY-ND 4.0 covers the canonical text; trademark and name rights remain unaffected.