Layer 5
Operational equipment: what you take for granted about yourself
Languages, jargon, domains where you move with ease and zones where you declare ignorance. What a good interlocutor takes for granted about you before starting to talk.
What it captures
Three blocks. Languages: with separate CEFR levels for read, write, and speak. Jargon: what you master (the AI can use it without paraphrasing) and the allergic kind (it actively avoids it). Competence and ignorance: domains where the AI can assume your level without explaining the basics, and domains where it explains from the ground up and never assumes prior context.
Why it matters
Without this layer, the AI defaults to a middle level: not so basic it insults your intelligence, not so technical it leaves you behind. That sounds reasonable until you find yourself spending half the conversation correcting the level: "no, I know that, go deeper" / "no, not that, don't talk to me like a beginner". This layer eliminates that dance.
How it is built
Three paths. One: the dedicated conversational module, which through ten turns with the AI maps your languages, jargon, and competencies. Two: you declare it from the dashboard, thirty seconds per item. Three: the AI detects candidates in your real conversations and sends them to the inbox for you to confirm.
The three areas
Languages
Separate CEFR levels for read, write, and speak. Because they almost never match.
Jargon
What you master (the AI uses it without paraphrasing) and the allergic kind (it actively avoids it).
Competence / Ignorance
Domains where the AI assumes your level and domains where it explains from the ground up.
Privacy and plan
Available only in the Professional plan. Low sensitivity in all categories. Soft delete throughout. The block is injected with a 250-token cap, prioritized by mention_count and pinned.
Frequently asked
- What if my English level improves?
- You edit it. Independent CEFR for read, write, and speak — because they rarely match. Update them whenever you have real evidence, no ceremony.
- How granular can a competence domain be?
- As granular as you want: "TypeScript", "TypeScript with React Server Components", "acoustic phonetics of Peninsular Spanish". Three depth levels: expert, working, surface. The more specific, the better the AI calibrates.
- Do allergic jargon entries apply always or only in some contexts?
- Always, unless you ask for a literal quote. If you declare yourself allergic to "synergy", the AI avoids it in replies, but if you ask to translate a text that contains it, it will translate it.