Adam Jacob shrunk his System Initiative team from 18 people to 5 and shipped Swamp 900 times in four weeks. That is the premise of this conversation, and it is not a vanity metric. It is a structural argument: AI agents do not just accelerate development, they collapse the organizational assumptions underneath it.
The most technically dense parts of this episode are not about the tools. Jacob argues that domain-driven design and software architecture now matter more than code-writing ability, because agents write the code. He also revived User Acceptance Testing from the 1990s as a quality gate, and he refuses to accept pull requests to Swamp, ever. The live demo where Stacoviak points Swamp at a Proxmox box and watches it generate its own automation is described in enough detail to be worth hearing firsthand.
Read the full transcript or listen to understand why Jacob thinks the bottleneck has moved upstream, from implementation to intent. The question this episode forces is not whether AI can write code. It is whether your architecture is legible enough for an agent to act on.
[READ ORIGINAL →]