Summary
Cole Medin and Raasmus host Brian, founder of BMAD (Brain Multi-Agent Design), for an extended conversation about AI coding frameworks, the future of software engineering, and how teams can avoid what Brian calls the ‘slop apocalypse.’ BMAD, now over a year old, was built after Brian noticed that even strong engineers were rejecting agentic coding loops as unreliable — the framework was designed to give agents structured planning artifacts (PRDs, specs, task lists) so they could produce consistent, maintainable output instead of accumulating context pollution.
The ‘slop apocalypse’ is Brian’s term for a failure mode that doesn’t manifest as broken code but as progressively more expensive and slower agent cycles — compounding token spend, increasing rewrites, and degrading coherence as agents operate on ambiguous or bloated context. His prescription is front-loading the work: thorough specification before any code generation, treating the spec pipeline as an engineering artifact in its own right. He also introduces the ‘human on the loop’ framing — engineers maintaining strategic oversight without being in the prompt-response loop for every step.
Brian and the hosts discuss how engineering roles are evolving toward ‘systems thinking’ — designing the full company build loop, not just shipping individual features. This creates a new risk: product managers or non-engineers writing specs that embed hidden architectural decisions before engineers review them. The conversation also touches on BMAD’s enterprise adoption work, the irreplaceable value of specialized roles even as AI accelerates output, and upcoming changes to the framework itself.
📺 Source: Cole Medin · Published August 25, 2026
🏷️ Format: Podcast
