Reraise holds your world in four layers, and any AI model can read them without owning them.

The four layers build on each other.

• Your personal world. Cards, sources, experience, assumptions. What you know, what you have lived through, whom you listen to.
• Your decision world. Rules, priorities, outcomes. How you decide, built from the cards of the first layer.
• The board. Where a situation becomes visible and you decide. It draws on both layers below.
• The platform. Worlds connected by choice, with their origin intact. Single cards travel; your world stays yours.

The first layer is your personal world: cards, sources, experience and assumptions.

Everything starts with the card: one structured piece of your world, with its origin attached. It can hold a fact with its source, a claim someone made, something you lived through, or an assumption you work from. How cards keep proven, said and believed apart is on Cards.

A world is every card that matters to you in one context.

A world is all the cards, connections, experiences and premises that matter to you in one context: a slice of reality seen from your angle. A gold market you follow. Your health. A property market you know.

One card can live in several worlds at once. It is not copied. It is used in context, because the same card can mean something different in each world.

Sources are judged on several measures, never on one score.

It is not enough to know that an analyst said something. What matters is how often that analyst was right. By design, Reraise keeps these measures apart: accuracy, reliability, expertise, agreement with your world, track record. One source is accurate but sees the world very differently. Another is accurate and agrees with you. A third is often wrong, and still valuable as the counter view. A single credibility score would hide exactly that.

This is designed, not built. Today every card records where it comes from; the track record itself is listed at the end of this page.

The second layer is your decision world: rules, priorities and outcomes.

The next level is not more knowledge. It is decisions. Knowledge says: I know the gold market. Decision knowledge says: under these conditions, I add to my position, and I would act.

A decision card holds:

• the situation and the trigger,
• the conditions and the known facts,
• your priorities and the action you prefer,
• the alternatives and the deal breakers,
• your reasoning and the cards you used,
• and later, the outcome.

The card format fixes these sections; until the outcome is known, it reads open. No training, no fine-tuning: a decision is first simply made explicit and retrievable. So one day your AI will not only ask what you know. It will ask which of your rules apply here, and how you would probably judge it.

The loop makes a world more experienced, not just bigger.

It runs in a loop. Experience becomes a card. Cards form a world. The world becomes a situation on the board. Analysis. Decision. Action. Result. And the result is new experience.

A decision that worked becomes relevant again. An assumption that keeps failing loses weight. Your world does not just get bigger. It gets more experienced. The decision card already holds its outcome; letting outcomes shift weights is part of the track record still to come.

An AI fed only your own decisions becomes an echo chamber.

The risk is real. If you feed your AI only your own decisions, you reward agreement. You get an echo chamber, and doubt gets trained away.

Doubt survives in one place only: where every decision also records how it turned out, including where you were wrong.

That is why the outcome is a fixed section of every decision card, not an optional note. And it is why a source that often disagrees with you is kept, not filtered out.

The third layer is the board, where a situation becomes visible and you decide.

Cards store what you know, the interface delivers context, the model writes answers. You still need a place where your world turns into a situation you can see and work with.

On the board, separate cards become one picture: the players, their interests, their moves, and the gap between what they say and what they do. You and your AI agents sit at it together. By design, agents are full players: they lay cards, make cards and draw connections. Only you set what you believe.

The board app works today on a developer machine with an example world. A clickable demo with example data is at the board demo. How fields are laid out and built is on Board and fields.

The fourth layer is the platform: worlds connected by choice, with their origin intact.

Every world is private by default. Single cards are shared on purpose. What is publicly findable can be shared; your conclusions and your beliefs never travel.

What a society knows exists only as scattered pieces held by many people.1 One person holds the original source, another the counter-evidence, a third a lived experience. Connected card by card, they form a structure no single person could build. There is no central truth and there are no editors.

The origin stays on every card, so copies are counted back to their source: ten thousand copies of one wire story are one source.2 The first exchange is at the trade page (beta), where public cards are brought in and taken. The rules are on Trading rules.

Your AI reaches your world through one interface that does not depend on the model.

You never paste your world into a chat again. An agent simply asks: what is relevant for this situation? Reraise answers with a context package: the active world, the relevant cards, their relations, your decisions, the sources and their origin. The model processes it. Your world stays outside the model.

One provider's model today, another's tomorrow, a local model, your own agents: all work on the same world. Reraise does not replace your agents. It gives them the layer that is missing: persistent knowledge and persistent decision logic.

In the tool that exists today, agents reach it through an API and an MCP server, the open standard through which an AI agent uses tools. It runs on a developer machine with an example world and is not released yet; the code goes public with the first trade. How to connect is on Connect.

Cloud models may be used, and the trade-off is stated openly.

Hybrid is the default. A card you allow to a cloud model goes to it as it is, so the model can work with it. Every such use is logged. One switch locks cloud use off entirely.

The trade-off is plain: cloud models are often stronger and need no hardware of your own, but a card sent to one leaves your machine for that request. Your stances, what you believe about a card, stay on your device either way.