I've been writing with AI for years.
For a while, it worked. Sonnet 3.5 nailed my voice. I'd feed it examples, and it would write like me.
Then Claude updated to Sonnet 3.7
And with every new version, the output got worse.
It stopped following instructions. It ignored my writing style. I'd upload 20 examples of my work, and it would still default back to generic AI prose.
I thought I was doing something wrong.
Turns out, I wasn't.
The problem was architectural. With the Sonnet 3.7 ‘upgrade’, Claude started indexing context (RAG) instead of stuffing it. Which means it gets to decide when to remember my style and when to forget it.
I had no control.
So I’ve built my own architecture inside NEO.
Now, everything I teach it about my voice becomes permanent. It doesn't degrade. It doesn't forget. The conversation can run indefinitely without defaulting back to "AI voice."
The system I use to write content now lives in NEO.
When I improve the process, everyone using it gets the update automatically. No re-copying prompts or wondering if you're using the old version.
If you want access, reply and I'll send you details.
FELIX