AI-tasks.app

If you see this page, you probably got here through my LinkedIn post, which said:

To experiment with AI, I decided to build a small personal project: an app that pulls from Gmail and Calendar to build a daily todo list. I estimated it would take me 4-6 weeks to build and deploy.

Before starting, I tried a different path: I gave Claude Code half a page of hastily written requirements and asked it to build the whole thing.

The Result: Claude crunched for 20 minutes.

The Debugging: It took another hour to clean up bugs (Claude did all the heavy lifting).

Cost: about $28 in Anthropic credits

I did not write a single line of code.

You can see the live app here: ai-tasks.app and the code here (NOTE: if you want to use it you'll go through some scary warnings from Google, since the app is not verified by Google).

Bottom line: We've hit a tipping point. Claude already performs like a solid full-stack dev. The role of the engineer is shifting from writing lines of code to managing "architectural intent."

I built the app only as an experiment and learning exercise, I had no intention of promoting it. It started using too many Gemini tokens ($$), so I shut it down. You can still take the code from the github repo and roll your own.