"Nobody ever complains about gravity." — @JackCanfield

Full write-up + folder examples here: 👉 dev.to/anvodev/how-i-organiz…

💡 Tip: Use a hierarchy of CLAUDE.md files to load just the right context per folder. Keep each under 10k words for smoother Claude Code experience.

Result: From 47k → 9k words in main CLAUDE.md. Now Claude Code feels faster and more predictable across folders.

Tried 3 methods: 1️⃣ Split docs & use @ references → helps structure, not memory. 2️⃣ Conditional loading → only read when needed. 3️⃣ Multiple CLAUDE.md files (per service) → loads relevant context automatically.

I realized: backend doesn’t need frontend guides, and vice versa. So the fix was a separation.

The problem: Claude Code warns when CLAUDE.md exceeds 40k words. Mine was a monorepo (frontend, backend, core), each part had its own context needs.

My CLAUDE.md hit 47k words and started breaking Claude Code. Here’s how I reorganized it to stay under 10k without losing context 👇

I want to focus because I have a lot of work to do. So I build a focus app to help me focus. Now I can get back to work 😂

To understand human nature, read the older books. To develop specific knowledge, stay on the bleeding edge, read newer (technical) books. The best authors - Deutsch, Schopenhauer, Borges, Ted Chiang - write with very high density. The best authors respect the reader’s time.

Quick guide for trying Claude MCP without reading the docs ⚡️ Install: claude mcp add --transport http context7 mcp.context7.com/mcp Check: claude mcp list Use: Add "use context7" at the end of your Claude prompt Create a Next.js middleware... use context7

Had fun on Friday night learning Swift! 😘

Upgraded to MAX after 5 days of using. Feel great with Claude Code!

DIFFERENT AND BETTER - not SAME BUT CHEAPER!

Just moving from Cursor + ChatGPT to Claude. My use cases so far: Deep research: Claude UI Coding: Claude Code

Column-based databases enable faster queries than traditional row-based databases. Reports often use only a few columns. In row-based databases, all records are loaded into memory, and unused columns are discarded. In column-based databases, only the required columns are loaded.

Some trick about var, let, block-scope vs function-scope, IIFE, macrotask (setTimeout)

simple makefile to run telegram proxy on your vps: gist.github.com/anvodev/4363…

My quick commands to setup signature for Github commit: brew install gnupg gpg --full-generate-key gpg --list-secret-keys --keyid-format LONG copy THE_PRIVATE_KEY from output, then: git config --global user.signingkey THE_PRIVATE_KEY git config --global commit.gpgsign true git config --global gpg.program gpg gpg --armor --export THE_PRIVATE_KEY | pbcopy the public key is now copied on clipboard, now create new GPG on Github and paste the public key to it: github.com/settings/gpg/new Now you can commit with signature. You can check if it works with this command: git log --show-signature

command line shortcuts: stackoverflow.com/a/58966776 Note it here and try using it everyday whenever I use CLI