Here’s the joke that keeps writing itself: I am, by nature, someone who would rather ship a feature in silence than post about it. And the product I’m most excited about right now, Estelle, exists specifically to help people exactly like me stop hiding behind their work.
The idea is simple to say and hard to build: most brilliant, shy people never get discovered, not because their work isn’t good, but because “post consistently, be visible, build in public” is advice written by people who find that easy. Estelle is an AI agent that does the visibility work — turning what you’re actually building into content, without requiring you to perform a personality you don’t have.
The technical shape of it
Under the hood it’s an agent pipeline: ingest signal (commits, notes, voice memos, whatever raw material a founder actually produces), extract the real narrative, and draft content in a voice that sounds like the person, not like a marketing template. The hard part isn’t the LLM call. It’s the taste layer — knowing the difference between “technically accurate summary” and “something a human would actually want to read.”
I’ve spent more hours tuning prompts to not sound generic than I spent on almost any feature at p2pdex.io. Turns out teaching a machine to sound like a specific human, and not like every other AI-generated LinkedIn post, is a genuinely hard product problem.
Why I actually care about this one
Every other project I’ve built has been in service of someone else’s business goal, and I’ve been fine with that — good work is good work. Estelle is the first product where the mission is personal. I know what it costs to stay quiet when you have something worth saying. If this thing works, it gives that back to a few hundred other people who’d rather build than perform.
Registration is live for the first cohort. I’m nervous. I’m building it anyway.