Knowledge became cheap.
Three things did not.
Machines drove the price of knowing things to nearly zero. What did not follow it down: knowing yourself accurately, governing yourself, and composing something from those two that no one else can copy.
The discipline that covers all three already has a name. StudyBrave makes it operational — as protocols. Named, versioned, evidence-linked, and open enough that you can take them and never come back.
Three layers, in order
You cannot govern what you have not mapped. You cannot build what you cannot govern. The order is not a preference.
Control
Can I predict and govern myself?
Create
What am I building that is mine?
Take them and leave
Every protocol publishes its full contents: the procedure, the agent definition that runs it, the schema of what it produces, and an evidence file that states what the protocol does not know.
Download the agent, paste it into whatever model you already use, and run it there. Nothing is withheld to make you sign up. The recipe is free; only the continuity is worth paying for, and that is not on sale yet.
- Every claim carries its evidence
- Each protocol ships with a file rating the support behind every design decision — strong, moderate, or weak. Where the support is weak, it says weak. The build refuses to compile a protocol without one.
- Nothing here is a diagnosis
- These are not assessments, tests, or clinical instruments, and none of them has been validated as one. The components rest on published work. The assemblies are untested, and say so.
- Adversary mode
- Every agent has a flag that stops it accepting your claims. It asks for evidence, names contradictions, and refuses the flattering reading. It is off by default.
There is no database of you. Not encrypted — absent.
Protocols run in your browser with your own API key. Your map, your record, and your calibration history stay on your machine. The server stores only what you deliberately type into a form — a newsletter address, or a contribution you sent to be read. Nothing is collected by watching you read.
During a run, your context goes to the model provider you chose, in plaintext, because that is what inference requires. Encryption protects storage, not inference. The security page states exactly what can and cannot be seen, and by whom.
One protocol every two weeks
The full procedure, the agent that runs it, and the evidence behind it. No announcements, no roundups, no course.