Mark LlobreraOptimizing my Eleventy build from 4 minutes to 45 seconds turned out to be a roundabout way to think hard about LLM use in my personal work: https://markllobrera.com/posts/optimizing-my-eleventy-build-an-ai-use-note/
Max Fenton@markllobrera This is an ideal blog post and I love the idea of doing the coding by hand. One idea would be that maybe something could do a one-time sizes calculation and store those values in a lookup table or asset media file. Should be enough For browser lazyloading and repaints.
@maxfenton Oh! That’s a real interesting idea
@markllobrera I'm trying to have the Claude Code stuff build tools (like you describe) and think a lot more about pre-caching/caching/lookups that would be too complex of a script for me to write but zero time for that tool to write once.
@maxfenton I think you’re the person who unlocked this lane for me, looking at LLMs as a tool for building…specialized tools
@markllobrera Also: SO much to talk about with you at some point about Claude files. I've got a token-hygiene rule now that audits project level .claude/CLAUDE.md files that use @-level includes based on CMS, filetypes, etc. (e.g. is there ddev? backstop? a config/project/project.yaml file?)
@maxfenton You know where to find me