8 works with eight kinds of human knowledge.
8 shows you which ones you and the AI are using — and which you may not be reaching.
Then it runs a learning loop that turns each reading into a sharper interaction, and in time, work you can inspect before you trust.
The empty rows are not missing data. They are potential insights.
The model draws on established distinctions in how knowledge is represented and used. The neural-basis paper included with 8 sets out the supporting evidence and its limits.
Facts and procedures are often the easiest knowledge to put into words. Other kinds — systemic Why, context-dependent When, embodied Body, episodic Story — can be harder to reach without a prompt. 8 provides that prompt.
Not a personality type. Not a label. A live picture of which kinds of knowledge you and AI actually use, right now, in a conversation you are already having.
The eight knowledge types provide a working read: what is present, what may be missing, what you are relying on most, and where an assumption needs checking. The loop is what you do with it — you scope the task around the gap, brief the model to cover it, and check whether the output earned its conclusions. Seeing and doing, on one substrate.
Five commands take a rough intention to a result you can check before you rely on it — scope, assess, brief, review, and adjust. New in v0.6, brief drafts a usable prompt in the same turn and labels the judgment points: Assumption (what 8 inferred), Required from user (facts only you can supply), and Human-only evidence (first-hand input the model shouldn't fake). You get a working prompt fast, and an honest list of what to check. The point isn't to remove your judgment — it's to make the draft inspectable enough for judgment to happen.
When praise escalates and nothing new gets checked, the work feels finished before it has been tested. 8 flags agreement that outpaces evidence and asks for the counterexample before you adopt it. Related: agreement and evidence →
Paste the prompt into your AI assistant (Claude, Gemini, ChatGPT) and start talking. The 8 Knowledge Types light up as you speak.
As you talk, 8 tracks two signals quietly. Friction (ƒ) tells you when the conversation is grinding — or gliding, with agreement outpacing evidence. Both are failure modes; 8 watches for each. Direction (Δ) tells you whether you are moving toward knowledge that is harder to reach or easier.
After a few minutes, type scan to see where you have been. Type trace to see the shape of the whole conversation. Type challenge any time the flow feels too smooth: the AI must produce the strongest counterexample to whatever was just agreed and rule honestly — breaks, holds, or needs evidence. The commands are simple. The patterns are yours.
When you're done, type save and the session becomes a durable record. Type synthesize [artifact] and the conversation becomes whatever you need: a summary, an email, a plan. Type focus [type] any time to move the conversation into a type you want to explore — like Body, Why, or Story.
Surprise — A conversation may be drawing on only three of the eight knowledge types. What else might become available?
Curiosity — What would happen if you went to the knowledge types you rarely visit? What is there?
Relief — A gap can become a place to explore, not a verdict about your thinking.
Agency — Once a pattern is visible, you can decide what to explore or change.
$96
The complete instrument: see your thinking through the eight knowledge types, and run the loop — scope, assess, brief, review, and adjust. Prompt, tutorial, reference sheet, neural basis paper, and 8 View. One purchase, unlimited use.