onBrand is a 29Court platform module that learns your voice: your tone, your word choice, your rhythm. Emails come out sounding like you wrote them. Summaries read like your summaries. Pushback lands like your pushback.
Sample verdict Paste one thing you wrote and get an instant taste of the verdict, right here in your browser. This quick check uses simple signals only. It is not your trained onBrand, and nothing you paste leaves this page.
Three steps. You bring the examples, we handle the rest.
Show us you at your best: emails, notes, posts, feedback, the hard no. Each one is a task plus how you actually handled it. Real beats polished, short is fine.
Your examples become your onBrand, a personal voice the AI writes through. Training fires on Together AI and needs an API key (your call, a few dollars). Nothing is faked, ever.
Every draft, summary, and reply comes out in your voice first. Pair it with 29Court doctrine rules and it stays honest too: your onBrand says how, the rules say what.
Write them here, in your browser. They stay on your machine until you choose to share them. When you hit 20, download the file and hand it to us.
Saved automatically in this browser as you go. The download uses the exact format our training pipeline expects.
Illustrative No trained model exists yet, so this is not a real before/after. These are real pairs from the demo dataset, showing the shape of what your onBrand produces once trained.
Where the machinery actually stands. No theater.
The full pipeline lives in the private 29court-brains repo under projects/onbrand/ (spec, scripts, real demo dataset).
Straight answers.
Style only: word choice, sentence length, tone, rhythm, how it opens and closes. It starts sounding like you on a good day.
Facts and knowledge. Twenty samples cannot teach a model what you know, and we do not pretend otherwise. If it did not know it before, your onBrand does not know it either.
You do. The samples on this page never leave your browser unless you download and share them. Your trained onBrand is yours: it is trained on your words, for your use, and nobody else gets it.
Doctrine rules are guardrails applied when the AI runs: cheap, editable, and they say what the AI must or must not do. onBrand is the voice underneath: how it says everything. They stack. Your onBrand writes it like you, the rules keep it honest.
A few dollars of Together AI credit for 20 samples. You bring the API key, the job runs server-side in minutes, and the finished voice is served as a hosted model you call like any chat model.