Most AI writing tools have the order backwards.
They generate first, then ask you to fix it. You end up iterating on garbage, hoping something real emerges from the slop.
--
I do a lot of case studies and technical content for deeply technical founders. My process has always been interview-first.
Talk to the customer. Get the transcript. Then throw it into an AI tool to synthesize.
The interview is where the real content lives. The AI is just the assembly step.
--
I came across Spiral recently, and something clicked.
They treat writing iterations like code commits. Small, discrete changes. You can see exactly what shifted between versions. You can track the evolution of an idea without losing where you started.
This is how engineers think. Change one thing. See what breaks. Commit. Repeat.
Most AI writing tools treat revision like editing a Google Doc: highlight, delete, retype. Context is a hard problem to solve, and they don't really understand my writing history. No valuable and easy to comprehend diffs. No understanding of what actually changed. No "getting" of my tone.
--
The other piece they get right is extraction before generation.
Great content isn't about prompts. It's about you, your specific perspective, your opinions, your way of framing problems.
So the tool interviews you first. Pulls the specifics out before writing anything.
This is already how I do case studies. Interview the customer. Transcribe. Synthesize. Spiral just packages that workflow into the tool itself.
--
The output of great AI content and slop looks identical. Same formatting. Same confident tone.
The difference is entirely in the process.
One commits small changes and extracts your perspective first. The other generates a wall of text and hopes you can polish it into something real.
Engineers know which approach scales.