Lumen learns what makes you you: your people, your places, your taste, your non‑negotiables. Then it hands every AI the same living context file. No more re‑explaining. No more confident guesses.
It feels like a game. Underneath, you’re building one structured file every AI you use can read.
Every answer is a tiny diff: one new fact, one new rule. “Ask me each time” is always a valid, honored answer. Lumen would rather pause than guess.
One small answer makes every AI you use a little more like talking to someone who already knows you.
A single portable Markdown file you fully own. Export, edit, or delete anytime. Scoped by default: work context goes to work tools, home to home. Nothing leaks where it doesn’t belong.
# Identity
name: "Alex Reyes"
pronouns: they/them
partner: "Sam" · # vegetarian, no cilantro
# People
kids:
- "June" (8, peanut allergy ⚠)
- "Theo" (5)
# Homes
- primary: Boulder, CO
- cabin: Estes Park
⚑ RULE: if "the house" is ambiguous, ASK.
# Trusted local pros
plumber: Vargas & Sons
electrician: Linde Electric
roofer: (not set)
# Taste
- tone: dry, concise, no exclamations
- coffee: cortado, oat milk
# Standing rules for models
⚑ Never invent specifics. Confirm before booking, buying, emailing.
⚑ High-stakes actions require a yes from me.
Context flows out to all of them. New facts flow back into the one file, automatically.
Wherever you have multiples (homes, cars, accounts, kids), Lumen stores an explicit ASK rule. Your AIs clarify before acting, instead of filling the gap with confidence.
Two minutes today. Every AI you ever connect, present and future, already knows who you’re talking about.