I replaced n8n with one I could actually trust

n8n was the obvious answer for a fifteen-step novel pipeline and a pile of agent loops. I built my own visual orchestrator anyway. The reason wasn't that n8n is weak. It's that when something breaks at step eleven of fifteen, I need to know exactly why, and a tool I wrote is the only one I can fully interrogate.

n8n was the obvious call for a fifteen-step novel pipeline. I built my own visual orchestrator instead, not because n8n is weak, but because step eleven of fifteen breaking in someone else's tool is a different problem than it breaking in yours.

https://www.jakelawrence.xyz/blog/the-in-house-n8n

The obvious answer for a fifteen-step-per-chapter novel pipeline was n8n. I built a visual orchestrator from scratch anyway. Not because n8n can't do it. Because when something breaks deep in a long agent loop, the only tool I can fully interrogate is one I wrote myself.

n8n was the obvious answer. Fifteen steps per chapter, a stack of agent loops, visual wiring already solved. I built my own orchestrator instead. Not because n8n is weak. Because the part that matters, when something breaks at step eleven of fifteen, is understanding exactly what state the system was in and why. That's a lot easier when the tool has no surface I didn't put there myself. https://www.jakelawrence.xyz/blog/the-in-house-n8n

n8n was the obvious answer for a fifteen-step novel pipeline. I built my own visual orchestrator instead. Not because n8n is weak. Because step eleven of fifteen breaking in my tool is a different problem than it breaking in someone else's.

https://www.jakelawrence.xyz/blog/the-in-house-n8n

n8n was the obvious call. I built my own visual orchestrator anyway. When something breaks at step eleven of fifteen inside a long agent loop, a tool I wrote is the only one I can fully interrogate.