I have 4 kids - Codex 1, Codex 2, Claude Code, and Gemini CLI.

Codexes have a special knack for development, I love them for that. Claude Code is the smartest kid I have; it acts as an orchestrator over all my builds and decides almost everything - my favorite child. Gemini was a mistake, but it does small UI tasks fine I guess.

I make "use" of all my kids. Multi Agent Memo was built as a shared brain in between these kids, so that they will know and acknowledge each other's work.

A couple of months ago I invested in SanDisk, Seagate, and Western Digital. Thank God I did, storage has been booming, and it's not gonna stop so soon.

Doesn't mean I should keep screenshotting my features to visually reference what I ship to LLMs. That's a hassle for me to cleanup. Agent Vision is a solution exactly for this. Agent Vision connects your IDE to your browser via Chrome Development Protocol, not only giving your LLMs ability to view your browser environment, but also your network tab, console logs-it's the entire deal.

I take complete ownership for the code I write inside and outside of work, which means testing every single flow takes a lot of time. Some flows are way too easy but have manual dependency since every single flow is browser based.

This resulted in the birth of Easy Sanity. Easy Sanity is a simple MCP tool that you setup in your IDE, you write down your test flows in excruciating detail in plain English and an AI agent just pulls them off for you.

Keeping shit simple, but useful.