Site policy
AI and this site
This site is still in its infancy. To move quickly without hiding how it was made, here is an honest split: generative AI helped with delivery mechanics (scaffolding, housekeeping, and repetitive implementation). The LSCSS methodology—ideas, judgement, experience, and technical detail—is mine, built from many years of professional and personal work with CSS in the wild, not from a model inventing a methodology.
What AI was used for here
Large language models and coding assistants were used to speed up work that is mostly mechanical when you already know what you want:
- Scaffolding and refactors. Initial page shells, repeated layout patterns, and refactors across many similar files.
- Link and consistency passes. Finding broken internal links, trailing-slash drift, and obvious copy inconsistencies between pages.
- Styles from an existing guide. Turning an agreed design direction (tokens, typography, spacing) into CSS and component-level rules—not inventing the system from thin air.
- Boilerplate and tooling glue. Build config, small scripts, and repetitive edits where the intent was already decided.
Those are acceleration layers. They do not replace deciding what LSCSS is, what trade-offs matter, or what goes on the page.
What was not outsourced to AI
The methodology itself—the ownership model, layer discipline, naming philosophy, how hacks and tokens fit, what to say about modern CSS, migration paths, and what counts as good enough in real delivery—is grounded in my lived experience shipping and rescuing CSS in teams and products. Models can phrase and structure; they do not substitute for years of doing the work, getting it wrong, and tightening the rules until they hold under pressure. If something on this site reads as generic or wrong, blame the author and the edit pass, not “the AI decided the methodology.”
How to read this site alongside AI
You do not need AI to use LSCSS. If you paste this content into a model anyway, treat the output as untrusted: verify against your stack, accessibility requirements, and your own constraints. The same applies to any third-party summary of a methodology—models flatten nuance.