The AI age is all about writing your own tools.
Recently, I've been getting obvious spam emails in my Gmail inbox, which are somehow bypassing Gmail's spam filter. These emails include attachments. I haven't clicked on the emails, because I can't trust that they aren't some form of attack that chains an exploit in Gmail's backend with an exploit in the browser or Gmail client.
But, I can just build my own email fetcher via the Gmail API, in about 5 minutes using an AI agent, and inspect these suspicious emails as safe plain text files.
Or, if I wanted, I could spend a day on this and add a webhook that kicks off a lambda that looks for suspicious patterns, and removes emails from my inbox before I even see them. Essentially, I could run my own spam filter against my Gmail inbox. This is something I'd never ever consider before AI agents, but now... it's trivial.