Layer 2
Negative space: what you don't want
An AI that doesn't know what you don't want ends up doing everything you didn't want. Pet peeves, disliked phrases, living antimodels, taboos, decisions you'd never delegate. Negative space is the silhouette around your profile.
What it captures
Four canonical categories: taboos (topics you'd rather not discuss with an AI), pet peeves (things that'd push your buttons if it did them), disliked phrases (lines you don't want to hear — the classic "great question"), and antimodels (people or styles your AI must never imitate). Each veto is tagged with severity: low, medium, or high.
Why it matters
It's harder to instruct someone on what not to do than on what to do — but also more useful. A single "don't open every reply with great question" cuts a hundred future hours of friction. And that's only the visible layer: negative space also intercepts the heavy stuff (don't bring up my divorce, don't suggest medical decisions, don't channel Tony Robbins).
How it's built
Three paths. One: the conversational Proust module, which asks indirectly and takes notes. Two: you add vetoes manually as they come to mind — from the dashboard, in thirty seconds. Three: the AI detects candidates in your conversations and routes them to the inbox for your confirmation (high-severity ones never inject without your green light).
The four subcategories
Taboos
Topics you don't want to touch with the AI, even in passing. Things you live with pain, closed decisions, intimate areas you'd rather keep separate from machine conversation.
Pet peeves
Those small things that'd drive you up the wall if it did them. Stylistically minor, behaviorally decisive. The line between tolerating the AI and embracing it.
Disliked phrases
Filler phrases that have crept into corporate-AI talk and you'd rather not hear. More restrictive than pet peeves: this is about specific words, not behaviors.
Antimodels
People, styles, or characters the AI must NOT imitate. The negative image of "speak like X": here you say "NEVER speak like Y".
Privacy and plan
Available from the Premium plan up. Edit, pin or remove any veto at any time (soft delete; nothing is lost permanently without your action). The block injects into the AI before your cognitive portrait, capped at 250 tokens and prioritized by severity.
Frequently asked
- What if I add something and later change my mind?
- Unpin or remove — soft delete, recoverable. Negative space is built to evolve with you, not to lock you into who you were two years ago.
- Can vetoes contradict each other?
- They can, and it's not a problem. Severity tags resolve conflicts: if two vetoes clash in a specific case, the high-severity one wins. If it's a tie, the AI asks before acting.
- Does it export with my portable profile?
- Yes, opt-in via `?include=negative_space` on the Professional plan portable JSON. Nothing exports unless you mark it.