TB: Information and Technological Evolution (Potter)

Reading Brian Arthur through an information-theory lens, Potter argues that the central engineering problem of inventing technology is search over an exponentially large space — and the trick that makes this feasible is modularity. Complex artifacts are not assembled from primitives all at once; they are built up hierarchically from stable subassemblies, each of which compresses the search problem above it. The framing turns "how does new technology happen?" into "how do you navigate large, complicated search spaces efficiently?"

• John's annotation in Blogs.org: "How do you navigate large, complicated search spaces." This is the article's organizing question.
• The space of possible logic functions is astronomically large — even a small 8-bit adder occupies roughly one out of 10^177,554 functions with the same inputs/outputs. Random search is hopeless.
• Hierarchical construction (Hora-style: stable subassemblies, then assemblies of those) collapses the cost: sequential gate-by-gate construction can find a working circuit in roughly 450,000 attempts, while attempting multi-gate combinations simultaneously requires billions.
• Information theory quantifies how much each successful step shrinks the remaining search space; entropy is maximized when outcomes are equally plausible, so informative stepping-stones are the ones with non-trivial outcome distributions.
• Technologies that specify clear intermediate goals provide stepping stones that make the combinatorial problem tractable.
• Brian Arthur's claim, recast in Potter's terms: "Technology forms a network of elements in which novel elements are continually constructed from existing ones."
• The same mechanism explains why knowing the right intermediate goals matters more than raw exploration capacity. Industries that lack legible intermediate goals tend to stall.

• "Complex circuits are built up from simpler technologies, the way Hora's watches are built from stable subassemblies."
• "When one outcome is very likely, you learn much less on each attempt, because you mostly get the outcome you already knew was likely."
• "Technology forms a network of elements in which novel elements are continually constructed from existing ones." — Brian Arthur, quoted by Potter

The piece reframes "technological progress" as a tractable computational problem under the right structural conditions, not a mystery. The information-theoretic framing is the same machinery that lets one reason about experimental design in physics: maximize expected information per trial. Direct connection to John's working interest in Generator/Filter-style search strategies in research, and to the Bitter Lesson's claim that what scales is search and learning, not handcrafted shortcuts.

• How Long Do We Wait for New Inventions? (Potter) — same author, complementary thesis on when inventions arrive
• The Bitter Lesson (Sutton) — convergent claim: search is the thing that scales
• Physics of Energy (vault) — adjacent reasoning about constraints

Information and Technological Evolution — Brian Potter, Construction Physics (Substack), 2026-04-02