Lead Architect Carlos Kelkboom shares his rules for coding agents and how great engineers architect software with AI. One of the great ones is: he throws away generated code the moment he doesn't understand it. He breaks down why coding agents fail at iterative changes, why his main agent never touches files, and how he scaffolds projects for one-shot execution.
In this video, we cover:
This episode is for software engineers, tech leads, and system architects who want to move beyond "vibe coding" and build maintainable, deterministic production software with AI agents.
Connect with Carlos:
https://www.linkedin.com/in/carlos-k-4016b8225
Timestamps:
00:00:00 - Why Carlos Throws Away Working AI Code
00:00:53 - Why Pull Requests and Team Collaboration Are Dying
00:03:50 - The Trap of Starting Projects You Never Finish
00:06:06 - Escaping the Solo AI Developer Bubble
00:11:54 - How to Scaffold Code Before Using AI Agents
00:15:54 - Learning New Programming Languages With AI Agents
00:21:14 - Why We Must Get Rid of the Heap
00:28:32 - Building a Shared Language Across Engineering Teams
00:31:31 - Vibe Coding vs Deep Software Engineering
00:40:12 - Code Is Cheap: Git Worktrees for Parallel Agents
00:41:45 - How Deep to Scaffold Before the Agent Runs
00:45:09 - Democratizing Software, Missing Juniors, and the AI Bubble
00:53:33 - Designing a Graph Harness for AI Agent Context
00:58:18 - Why His Main Agent Never Touches Files
01:03:07 - Throw Away Code You Don't Understand
01:06:25 - Rethinking Code Reviews and Mutation Testing for AI