Meet Chloe, the AI assistant that now lives inside SelfSM.
She exists because I failed at something simpler. I kept trying to make onboarding easy: cut steps, redesigned the flow, moved things around. Every version worked for a while, until the product grew again, the flow gained new screens, and new users were back to guessing where to start.
The fix turned out to be the most trivial idea on the list, which is probably why I tried it last: stop simplifying the map, give people a guide.
Chloe takes you from an empty workspace to a complete brand setup, weekly content plan, and first drafts in one conversation. She asks what you do, sets up the brand, and interviews you for actual experiences and opinions before generating anything. Feed her your past posts and she'll pick up your voice from them.
Two things mattered to me. First, she has access to the whole workflow: voice, strategy, weekly plans, publishing. That's 25 tools connected to the same product actions available through the UI, so most things you can do by clicking, you can now ask for in plain language. Second, when the required material is missing, she says so. I tested this on purpose: asked for a post with nothing behind it, and she refused to write it, flat out, instead of filling the gap with something generic. Publishing still requires an explicit confirmation. Chloe cannot ship anything on her own.
The old onboarding wizard stays as a fallback. I don't know yet whether new users will prefer talking to Chloe or clicking through the screens.
Hey Product Hunt, Vladimir here, solo dev on this.
SelfSM writes content in your client's voice instead of a generic AI one. You set up a voice profile per client: phrasing, words they never use, the facts about them that are real. Every post gets generated against that profile, not against a blank prompt.
The part I care about most is the fact gate. Most tools will happily put a number or a credential into a post that nobody gave them, and you find out when the client does. Here the model can't insert a figure unless it came from you. It's enforced at generation, not just asked for in the prompt.
Also in: pull fresh news on a topic and write a post straight from the article, in the voice you already set up. No copying links into a separate tab.
It doesn't schedule and doesn't post anything for you. You approve every post.
If you write for clients, tell me where this would break for you. That's more useful to me today than an upvote.
The day my own AI invented client metrics, and what I built because of it
Early beta, a test draft came back with a revenue number the client never mentioned. Looked plausible. Was completely made up.
That bug became the core feature: every claim in a draft now has to trace back to the client's fact base, or it gets blocked before anyone sees it. No fact, no claim.
If you've used AI for client content, how do you catch fabricated details before the client does?