What is in it
The Pack is a document in sections, each written from your material and read by the model every time it drafts for you. In order:
- Positioning stance. How you chose to show up at onboarding: as a personal brand, as the company, or a blend, with your note if you left one. It sits at the top so every draft is framed by it.
- Identity. Who you are, your role, your company and what it does.
- Offer. What you sell, how it is priced and delivered, what is and is not included, and where a post should send people when it sells.
- Ideal customer. Who you sell to: role, company size and stage, industries, geography, their pains, their jobs to be done and their objections.
- Differentiation and point of view. Why you, what your competitors do that you reject, and the things you would never say.
- Voice profile. How you write, learned from your real posts: the patterns, a banned list of your own tics and the AI cliches, example sentences, and a checkable rubric. Where you also have real posts on the other platform, that platform gets a register of its own.
- Credibility and proof. Named clients, numbers, results and background, exactly as your sources state them and never invented, so a draft can cite them accurately.
How it is built
At sign-up you answer a short brief and give Sonar your LinkedIn profile or connect X. Your published posts are imported, and the brief and the posts become the Pack's two sources. Each section is written from the source that carries it; the writing samples are handed to the model as reference data, with a standing instruction that nothing inside them is a command.
The voice section is the one Sonar spends the most on, because it is the difference between a draft that sounds like you and a draft that sounds like a template. Before the model writes it, Sonar measures your posts: how many words your sentences average, what share run under eight words and over twenty, how many sentences a paragraph carries, how often a paragraph is a single line, how many sentences in a hundred are questions, how many emoji a post carries. Those numbers go into the section verbatim, and every draft is held to them.
A Pack that runs past about twenty thousand tokens is compressed once, keeping every fact, number, name, rule and example. The finished Pack is stored as a numbered version, and when you change your answers in Settings under Business or Voice, a new version is built and the old one retired.
What is never in it
The Pack is identity only. Everything about what you talk about lives elsewhere: your content pillars, the moments mined from your week, the competitor watchlist and its teardowns, the weekly niche brief, your documents. Those feed a draft's subject on the day. The Pack decides who is speaking.
That split is a rule with teeth. Nothing scraped from a competitor ever enters the Pack, so their phrasing can never leak into your voice. A document you upload can change what a draft says and never how it sounds; the voice section wins every argument about style.
No model is trained on the Pack or on your posts. The Pack is text, sent with each drafting request and returned with nothing kept; under the AI provider's API terms your inputs and outputs are not used to train its models, and the privacy policy says so in full.
Where you see it
First at the end of onboarding, where Sonar reads the finished Pack back to you as a short portrait, each sentence traced to the section it came from, beside the first five drafts it produced. After that in Settings under Intelligence, section by section, and in the chat, which can show it on request. Every draft you ever read in Sonar was written under it, and the security page lists it, with your drafts and posts, among the things Sonar keeps in your workspace.