Discussion about this post

User's avatar
VibeQuant's avatar

"Domain expertise as moat" - 100% this.

Pawel Jozefiak's avatar

The observation about developers "firing multiple agents without even looking at the final code result" over Christmas - I saw this shift too. Something fundamental changed.

Your point about domain expertise being the moat is exactly right. I've been building with Claude Code for months now and the differentiator isn't prompting skill - it's knowing what you're building and why. The AI handles syntax, you handle strategy.

The Plan Mode tip is underrated. I use it constantly now for anything beyond trivial changes. Seeing Claude's thinking before execution has prevented so many "oops, that's not what I meant" moments.

On your prediction about "agent harnesses" - already seeing this. My personal agent system (Wiz) is essentially this: a framework for managing multiple specialized agents, not the agents themselves.

Wrote my honest assessment of Claude Code (what works, what doesn't) here: https://thoughts.jock.pl/p/claude-code-review-real-testing-vs-zapier-make-2026

No posts

Ready for more?