How we choose to operate: a compact compendium of our working philosophy in the age of AI.
Small system: 15 short articles in 3 categories
Categories are classical: Ethos (values), Praxis (methods), Telos (purpose).
Each article is an idea: a node in a graph, composed via hyperlinks.
Start anywhere.
Dig deeper in the bibliography.
This isn't theory but a practical mindset for the operator. It was discovered with serendipity while exploring software engineering, and battle-tested iteratively in our work, for well over a year. These principles have proven useful to us in most domains.
Our Ethos (values) start from human agency: because the tool is a lever, understanding remains the work, and the artifacts (results) we seek are boring enough to be intelligible without magic.
Telos (purpose) aims that stance toward augmented theory-building: systems should become intelligible, capability should compound, and intent should live in literate systems beside code, data, outputs, and context. Ultimately, we believe this may converge toward the new personal computer.
Praxis turns this into method: research for signal, go slow then fast, separate exploration from artifact, practice context discipline, use the generated buffet to select rather than accumulate, build composable boundaries, and write literate source.
A name for this territory might be: literate, composable augmentation.
Or, more explicitly: software practice as augmented theory-building.
That is: using computers, tools, and now AI to improve the human capacity to understand, design, build, and maintain systems, without pretending that the tools remove the need for understanding.
Most of the ideas collected here are not new. They belong to a long line of work on human-computer augmentation, programming as understanding, literate artifacts, composable systems, and disciplined simplicity.
This is our humble attempt to learn from that lineage and apply it to the way we think then make technology.
- Human augmentation: computers should increase human agency and capability, not bypass them (Licklider, Engelbart, Kay, Victor).
- Understanding: the real work is building a theory of the problem, system, and constraints (Naur, Brooks, Dijkstra).
- Literate systems: intent, rationale, structure, data, executable work, and outputs should share a readable computational medium (Knuth, Wolfram, Pérez, Granger, Howard).
- Composability: small, bounded, intelligible pieces should combine into larger capability (McIlroy, Thompson, Ritchie, Parnas, Kernighan, Pike, Hickey).
- Context discipline: context is an engineering material: selected, shaped, scoped, and delivered for a purpose.