About this collection
LLMs change the cost of producing code and prose, but they do not remove the cost of understanding, verification or ownership. This hub brings those concerns together with older software-engineering questions: what makes a change safe, when an abstraction clarifies a boundary, how dependencies expand a threat model, and why readable output can still be structurally wrong.
The selected path begins with maintainability and learning, then connects model-assisted work to security, framework design and philosophical accounts of AI. The full list is intentionally broader. It includes service objects, callbacks, language syntax, functional abstractions and adversarial protection for artists because the practical consequences of generative systems reach beyond code completion. These investigations do not assume that more automation is either progress or decline. They ask which evidence survives the demo, where review must remain human, and how a team can preserve a coherent system when plausible output becomes cheap.
Start here
- Good code after LLMs: maintainability was never just readability What maintainability means after LLM agents, and why explanation speed does not replace design integrity.
- If LLMs can write code, why learn programming from books? Whether programming concepts still matter in a world where LLMs and coding agents can generate production-grade code.
- Supply-chain attacks, npm, Laravel, and Ruby teams Npm and Laravel-style supply-chain attacks, and how Ruby and Rails teams can reduce risk in production, CI, and local development environments.
- APoSD and Clean Code from a Ruby point of view A Ruby-focused synthesis of John Ousterhout and Robert C. Martin on method length, comments, TDD, deep modules, and practical team policy.
- LLM evolution through a phenomenological lens LLM evolution read as a change in meaning, action, time, trust, and the lifeworld rather than as a brain, tool, or species metaphor.
All investigations
- Music discovery by conversation
- Ruby, Rails, COVID and the LLM wave
- APoSD and Clean Code from a Ruby point of view
- Go's semicolon insertion rule and why braces feel weird
- Callbacks, invariants, and durable events in Rails
- SICP ideas in practical Ruby
- Supply-chain attacks, npm, Laravel, and Ruby teams
- The Rails job that ran before checkout committed
- Brotli, gzip, and Propshaft in Rails
- Ruby, Go, Rust, and Zig for simple CLI apps
- On zygohistomorphic prepromorphisms
- LLM evolution through a phenomenological lens
- Good code after LLMs: maintainability was never just readability
- Optimistic locking for editing a CRUD app
- If LLMs can write code, why learn programming from books?
- sqlc, iBATIS, and SQL-first data access
- Propshaft versus Sprockets in Rails assets
- Private and protected methods in Ruby
- Service objects and Rails conventions
- Handling Server-Sent Events from a Rails backend
- Glaze, Nightshade, and adversarial protection for artists
- Relational algebra, SQL, and Arel
- SQLite, PostgreSQL, and the smallest useful database choice
Related main-site work
- Rails Dependency Pruner A review-oriented experiment that keeps static analysis, runtime evidence and rollback separate.
- EmbeddingGemma Inference Service for Rails and Kamal A small, versioned inference boundary for a production Rails search feature.
- The analogue programmer in the land of LLMs A main-site essay on authorship, tools and professional judgment.
- API Redux A maintained example of explicit contracts and operational design.