claudiacsf

Steve Yegge offering a peek into his 50-agent-1-human software factory. Field notes: Every time you think he's talking about a human, it's most likely an agent.

Here is how you build an agent that opens its own issues, fixes and closes them. Sampled from a real life loop observed in a European retailer - one of our customers - with 1500+ engineers.

We interviewed 2 engineering managers on how AI changed their jobs. Feel like this is a good read and might sound familiar to waterlogged EMs. Their closing advice? Knowing what not to work on.

TL;DR As a code quality platform, we're seeing first hand the changing processes inside dev teams. We noticed that the "meaningful" point of change is happening more and more at the local level, not at the repo level, where a team used to come together to review changes. With more teams waving past code reviews, the typical quality infrastructure that used to ensure code maintainability, security, reduce tech debt, etc, is breaking apart.

If you're a dev who cares about this, the new quality and security enforcement infrastructure needs to happen alongside the agents, locally. The tools to address this are still in early stages - meaning, a tool that isn't the same model reviewing itself - but they're starting to surface and it's worth keeping an eye on them.

Yes, please. More drones for good.

But I get triggered by the em dashes. :(

We wrote about how AI-assisted development changes the economics of code review and how good ole SAST does actually have a meaningful role to play today.