Building your AI Stack

Most people pick one AI tool and try to make it do everything. I get it. But after a few months of real daily use, I've landed on three. They don't overlap as much as you'd think.

I manage infrastructure for a large organization, so I don't have the luxury of experimenting with AI tools that can't survive a real workday.

Claude is my thinking partner. When I need to work through something, a strategy doc, a tricky script, a problem I can't quite frame yet, I go here. It's the one I talk to, not just at. But it's also my builder. Claude Code handles my coding projects. I use it as an MCP to pull data from other sources. And Cowork handles the stuff I don't want to think about anymore, like sending daily briefs and cleaning up my folders at the end of the day.

Gemini lives inside my Google stack, so it's become my operations layer. It proofreads my emails before I send them. NotebookLM lets me throw a pile of documents at it and get a conversational podcast back. That still feels like a party trick. But I keep going back to it, so maybe it isn't. It's not flashy. It's just there when I need it, which matters more than flashy.

Perplexity is my research reflex. When I used to open a browser tab and start Googling, I open Perplexity now. It's faster, the sources are right there, and I waste less time falling down rabbit holes. Mostly.

The thing nobody really talks about yet is the switching. Knowing which tool fits the moment. That's not a feature any of them advertise. It's a skill you build by actually using them on real work, not demos.

I'm not loyal to any of these. If something better shows up tomorrow, I'll try it. But right now this is the kit that's actually earning its place in my day.

Curious whether other people are settling into multi-tool workflows or going all-in on one.