Content in English, in this account. Español → @serudda_es

80% of slop UI comes from prompting "add a dropdown" when you meant a Combobox. built a POC: paste any UI screenshot → every component identified with standardized names from @uiguideline's research across 20 design systems. Identify every component name without memorizing a single one.

If Epstein managed to do what he did with the resources he had at the time... I don’t even want to imagine what he could have done today using AI.

I’m burning a lot of tokens with @openclaw and couldn’t find a native way to switch models. So I created a SKILL.md to quickly change OpenClaw’s thinking mode: eco mode → summaries, quick questions balanced mode → daily driver smart mode → complex work max mode → critical tasks github.com/serudda/switch-mo…

The uncertainty of watching the world move at 20x speed. Feeling like you can’t keep up. Wanting to, but not feeling good enough. The harder I push myself, the more the fire grows inside my head.

Action not only accelerates useful failure, but also generates information that paralysis can never produce.

I felt overloaded with ideas, most of them disappearing into the void. That’s why I build my own Second Brain: - Grounded in the Zettelkasten methodology - Capture, connect, and compound my thinking instead of losing it Check it out ↓ github.com/serudda/zettelkas…

A message to your past self: “You weren’t choosing for the rest of your life. You were choosing only for this season.” What you start today doesn’t force you to stick with it forever. You learn from it, and if it doesn’t work or you don’t like it, you simply choose a different path.

I tried @antigravity because of its generous PRO model limits. · But it’s a poorly made @cursor_ai clone. · The autocomplete is the slowest I’ve seen (it reminds me of @GitHubCopilot when it first launched). A couple of times, I had to switch back to Cursor because I couldn’t stand the Antigravity experience. (The worst part is that this is a product from the 4th richest company in the world.)

prompting claude when it makes mistakes

Conspiracy theory: @claudeai degrades its current models weeks before launching new ones. This makes the new model/version feel "smarter" than the previous one.