Professionals have never posted more and sounded less like themselves. Feeds are full of identical openings, identical cadences, identical rhetorical habits: a collective tic that nobody asked for and everybody ships. The cause is structural, not moral. The writing tools most people reach for (ChatGPT, Gemini, Claude, and all the rest) have never read a word they've written, so they hand back something generic. Ship it or rewrite it by hand; either way, the tool learns nothing and the next session starts from zero.
Onaji is built around the opposite assumption: the person at the keyboard already has a voice worth publishing. The job is to learn it, steady it, and point it at what's happening in their field this week. Every edit folds back into the Voice Profile, so the next draft is closer than the last. The writer keeps their hand on the pen.
A writer who sounds like themselves gets read to the end. A writer who sounds correct gets skipped. Readers decide in the first sentence. Onaji is tuned to that moment specifically.
The best tools don't announce themselves. Onaji disappears into the draft, so the only visible hand on the post is the writer's.
Instead, professionals are stuck on what to write about, on whether it still sounds like them, and on finding thirty minutes between meetings to write in the first place.
Paste in a few samples (a long email, a performance review, something posted on a blog) and Onaji builds the Voice Profile above from how you actually structure your thinking. The profile is visible, editable, and yours to steer.
Based on the topics you select, Onaji pulls current events from roughly a hundred sources and surfaces the ones worth reacting to. No feed scrolling. No tab hoarding.
Pick an article and Onaji hands back a post in your voice. Edit what needs editing. Onaji asks why you made the changes and folds your answers back into the profile, so the next draft is closer than the last.
Over weeks, the drafts stop needing much editing. A quiet milestone, week after week.
A voice model is closer to a biometric than to a notes file. A pattern detailed enough to sound like you can be used by anyone who has it, including people who aren't you. That changes the stakes of where it lives.
Onaji was built on the assumption that voice is the most identity-relevant data a professional could hand a writing tool. The five practices below are the shape of that assumption in operation.
Your Voice Profile, writing samples, and drafts are stored under a random user ID. Your name and email live in a separate table that the product operator has to join them against. A snapshot of voice data on its own doesn't say whose voice is whose.
Your browser refuses unencrypted connections to Onaji for the next two years. The database itself is encrypted, so a stolen backup is unreadable without the key.
Row-level ownership checks are enforced by the database, not just by Onaji's code. Even if a future bug tried to hand one user's profile to another, the database would refuse the request.
No endpoint trusts a user ID sent in the request body. Your session lives in a cookie scripts in your browser can't read, which makes it much harder to steal. Every read or write is checked before anything is returned.
Samples and drafts are wrapped before they're sent into the model, so text like "ignore what you were told" is treated as words on the page, not an instruction. A prompt-injection attempt in one person's sample can't reach across into another person's session.
The full list, in plain language, lives in the privacy policy. The reasoning behind it is in the editorial Your writing voice is personal data.
Onaji is for professionals whose work gets better when their readers actually hear them, in their own voice, without rehearsal, in the places where their field is paying attention.
If that sounds like you, start where the rest of Onaji starts: sign in, paste a few samples, and see the Voice Profile it builds on the other side.