Everyone is writing about agents these days, so I figured I’d write about agents too :)

This is a screenshot of Yellowduck, a tool I’ve been vibecoding for the last three months.

We basically dump everything we have at welley.social into it: repos, bug tracker, Figma, Notion, etc and 27 different agents work on top of it.

They investigate issues, review code, file bugs, fix bugs, join discussions and generally do a bunch of the boring work around software development.

One important distinction: Yellowduck itself was vibe-coded. Our products aren’t.

We use agents heavily in the development process: extensive code review, investigations, bug filing and bug fixing. But agent-written code still gets reviewed by both agents and humans.

For example, drop a bug or feature description into Slack and the agents can investigate it, fix it, file it for later or jump into the discussion.

The second screenshot is Bugsy, one of the agents, actually having a conversation in one of our product channels.

Lots of people are doing similar stuff already. But it’s been less than a year since claude/opus 4.5 came out last December and it’s already pretty damn obvious where this is going.

At some point, you’re going to look at most of the companies and realize that a ridiculous amount of the work is just agents talking to other agents.

All the boring stuff is going to disappear.