Examples
What changes when every AI you use reads the same file before it answers. You write your Creed once, each agent reads it before it replies, and it stays plain Markdown you own.
Builders
Your stack and standards, carried across every coding agent.
Plan in one tool, build in another
You sketch a service in ChatGPT, then build it in Claude Code. Both use your house style and your default stack. You paste nothing between them.
Shipped goals prune themselves
You tell an agent v1 is live. It proposes removing the old goal and promoting the next one, so your goals never rot into a graveyard.
It stops suggesting tools you banned
An ORM burned you once, so your file rules it out. No agent, in any repo, opens a pull request that adds it again.
Swap coding agents mid-task
You hit a limit in one agent and move to another. It already knows your commit style and PR rules, so nothing resets to defaults.
Teach one agent, fix them all
You correct Cursor once on how you log. It saves the rule, and Claude Code respects it weeks later without being told.
Your conventions in every new repo
Open a fresh repo and the agent scaffolds it your way: your router, your syntax, your defaults. There is nothing to copy over.
Writers
One voice, in every tool you write with.
Your voice survives the switch
Draft in one tool, edit in another. Both know your tense, your rhythm, and your rules, so the edit still sounds like you.
It drops the tics you keep cutting
You keep deleting the same filler opener. The agent proposes a rule, you accept, and it stops appearing in every tool.
Brand rules no agent forgets
Banned words, no exclamation marks, your spelling rules. Every draft from every tool comes back already obeying them.
Two client voices, kept apart
Warm for one client, clipped for another. The agent picks the right register from the first line and never blends them.
Jump tools when one gets worse
Your usual AI degrades after an update. You move to another mid-draft and it continues in your voice, with no re-briefing.
Formats it already knows
Your newsletter, your release notes, your pitch each have a shape. Any agent fills it in instead of inventing a new layout every time.
Researchers
The rules of your field, applied without a reminder.
Every agent cites your way
Draft, gather sources, and polish across three tools. All of them return citations in your required style, unprompted.
A rejected source stays rejected
You exclude a paper in one tool. Weeks later another leaves it out and notes why, because the exclusion lives in your file.
Explained at your level
You are an immunology PhD. Assistants skip the textbook basics and engage at your level, in your field and outside it.
It speaks your stats stack
You work in R, not Python. Every agent gives you code and examples in the tools you actually use.
Your methods, respected
Your methodological commitments are in your file, so no agent proposes an approach you have already ruled out.
Field terms it gets right
A word means something specific in your field. Every assistant uses it correctly instead of the popular-science version.
Operators
How you decide and communicate, read before any draft.
Every agent knows how you decide
Ask two tools for a recommendation. Both lead with the call, not a both-sides memo, because your file records how you decide.
Drafts in your board voice
Ask for a board email and it opens with the number that moved, the risk you watch, and the decision you need. It reads like you.
Advice that fits a lean team
Ask how to fix a slow queue and it proposes tooling before headcount, because your file says you stay lean this round.
Your hiring bar, across tools
One agent screens a resume, another writes interview questions. Both apply the same bar, because your file records how you hire.
Advice tuned to your risk bar
Recommendations match how much risk you take this year, because your file records your appetite instead of a generic default.
It protects your deep-work block
No agent schedules a meeting before noon, because your file blocks your mornings for focused work.
Ownership
Plain Markdown you own, portable across tools and providers.
Leave a vendor, keep everything
Cancel one AI and open another the same day. It already knows you, because your context never lived inside the subscription. Export it as Markdown anytime.
A new agent knows you instantly
Connect a brand-new tool, click Allow, and its first reply is already tuned to you. No setup chat, no priming prompt.
Fix it once, every agent updates
Tell one agent you go by Sam. The next morning every other tool opens with Sam too. One file, not four memories that forget separately.
Version history you own
Push your file to GitHub and see every change: a career switch, a closed goal, a move. Diff it, or roll back a bad edit.
Run it on a model you host
Export your file and feed it to a model on your own machine. The same context works with no vendor in the loop.
Back in seconds after months away
You step away for six months and come back. Your file is exactly as you left it, and a reconnected agent picks up where you stopped.
Boundaries
Lines an agent reads first and never crosses, long after you set them.
The topic AI never raises
After losing your father, you wrote one line: do not use him as an example. Months later, no assistant ever wanders into it.
No alcohol, ever, no awkwardness
You are sober. No assistant suggests a wine pairing or a bar, and you never have to explain why to a new tool.
Private stays private
Your wellbeing agent knows your diagnosis. Your work drafts never mention it, because the line is set once and every tool reads it.
Ask before the irreversible thing
Agents can draft and send for you, but never to your manager or investors. Those drafts stop and wait for your eyes first.
Limits move only when you approve
An agent thinks a rule about your kid is out of date. It proposes the change and waits. The boundary moves only when you say so.
It won't reopen a settled decision
You decided not to rebrand. When an agent drifts back to it, it stops, because your file marks the call as closed.
Everyday
The ordinary wins, for anyone.
Your kids' ages stay current
Your daughter turns five, an assistant proposes the update, and you approve it. Months later a different tool suggests age-five gifts on its own.
Never paste your bio again
The intro you used to paste into every tool is just there in your file. A new laptop or a new app changes nothing.
Meals that fit your diet and budget
Ask for a week of dinners and it lands pescatarian, dairy-light, and on budget. A new agent gets it right the first time.
Training that respects your injury
Ask any assistant for next week's runs. It caps the long run and works around your bad knee, because it read your goal and your schedule first.
Trips booked the way you fly
Ask any travel agent for a weekend away and it picks an aisle seat, skips the red-eye, and finds a hotel near transit, because your travel habits are in your file.
It remembers the dates that matter
Birthdays, anniversaries, and how you like to mark them live in your file. Any assistant plans ahead for them without you flagging the calendar.
Health and safety
Facts that have to hold in every tool, even ones with no medical context of their own.
Your allergy follows every assistant
You plan dinner with one AI, shop with another, and scale the recipe with a third. None suggest the shrimp, because all three read your allergy first.
Medications checked before advice
Ask about a cold remedy or a cheese board and the assistant flags the conflict with your medication instead of handing you the thing that harms you.
Built for how you actually read
You use a screen reader. Every agent describes charts in words and explains code in prose, without being told again in each new tool.
Plans that avoid your triggers
Your migraines come from strobing light and long screens. A scheduler adds breaks and an events agent skips the laser show, because both read how to help.
Plans broken into small steps
Long plans come back as short, ordered steps with one next action, because your file says that is how you work best.
It respects your sleep window
No assistant suggests a task past 10pm or schedules one before 7am, because your sleep window is in your file.