Prism: the right side of you, in every AI

Try it

Same question. A different hat.

Pick a hat and a question, then compare the answer with Prism to the same question with no Prism.

Marcus

The file the AI reads profile only

loading…

The answers below are prewritten for this demo, so editing the file won't change them.

Which hat to wear (his profile loads with every hat; a hat adds its voice)

▲ With Prism

Prism · Default profileReads the file above

Pick a question. This answer reads Marcus's file.

▼ The same question, no Prism

Plain AIDoesn't know you

The cold-start version: no file, no hat.

Top row reads Marcus's file; bottom row gets nothing.

The problem

You don't have a self online. You have strangers.

You pay for a handful of AI tools and teach each one the same things (your style, your work, what you're trying to do) every single time. Switch tools and you're a stranger again.

Your Claude-self

A stranger. Re-teach everything from scratch.

Your ChatGPT-self

Knows the half you told it last month. Forgets the rest.

Your Gemini-self

Another stranger. Same introductions, all over again.

How it works

Three things to say, in any chat.

`load prism`

load prism

Which hat? Investor · Work · Family · Tennis player

investor

Loaded your profile + Investor. Other hats stay sealed.

Saying “put on my investor hat” works too.

`save prism`

save prism

Save to Investor: “Never put more than 5% of savings in a single stock.” OK?

Saved.

Nothing is saved until you approve it.

`prism`

prism

Investor hat on. Latest saved: “Before any new position, write down what would make me sell.”

Which hat is on, and what was saved last.

You add Prism once, as a connector in your AI's settings.

Saving works in Claude and Gemini. On personal ChatGPT plans, Prism is read-only.

A normal Wednesday

What it looks like in your week.

Marcus's day. One profile. A hat for the side of him the moment needs. The same Prism in every AI he uses.

Wednesday: Marcus, with Prism

9:14 AM

Opens ChatGPT: “Put on my investor hat. Pressure-test this position for me.” ChatGPT goes cold and skeptical: unit economics, the downside, what would change his mind. Exactly how Marcus thinks about money.

11:40 AM

Switches to Claude, says load prism and picks Work. “Draft the opener for my board update.” Claude shifts to board-room voice: headline number first, one page, no filler. Same profile, different hat.

6:30 PM

Opens Gemini with his family hat. “My youngest wants to quit soccer mid-season. Help me talk to them.” Now the AI is warm and patient. No markets talk: the investor hat is sealed out of this chat. Gentle, human.

9:48 PM

In Claude on his phone, investor hat on, he says save prism. Claude drafts one line: “Never put more than 5% of savings in a single stock.” Marcus says OK, and it's saved to his investor hat ✓ approved. The next AI he wears that hat in already knows it.

One profile. A hat for each side of him. Claude, ChatGPT and Gemini. Nothing saved without his OK.

Sealed hats

Your investor hat never shows up at the dinner table.

One hat per chat. The AI gets your profile and the hat you put on. Nothing else.
Enforced by the Prism server, not the prompt. Ask for another hat's contents and the server refuses.
Switching is on purpose. Say “put on my family hat” and the investor hat comes off.
Sealed between chats. You can always see and edit your hats.

It's yours

Plain files you can read, edit, export and delete.

✓You can read every word. Your profile and each hat are plain text files.
✓Edit or delete any line. Including anything an AI saved.
✓Export them. Take your files with you, in a format any editor opens.

profile.md + one file per hat: Marcus's copy

# Default profile
style: direct · bullets · no hedge
home: a wife and two kids
work: a long career in finance

## hats
  · investor.md (cold, skeptical)
  · work.md      (board-room)
  · family.md    (warm)
  · tennis.md    (tennis player)

› readable. editable. exportable.