What the model does, and why it is not enough.
When you ask a language model what to do, it does something extraordinary. It takes everything anyone has ever written down that is roughly relevant to your question, and it gives you back the most statistically likely next sentence. This is a useful trick. Some of the time the most likely next sentence is also the right one and the model looks like a genius. Some of the time the most likely next sentence is fluent nonsense that sounds exactly like the right answer and the model looks like a genius then too. The model cannot tell the difference. That is not what it does.
The model has never raised a child. It has never lost a parent. It has never carried a business through a recession or watched a marriage fail or recovered from anything. It has read everything that was written about those experiences by people who did them, and a great deal more that was written by people who did not, and it has no way to weigh the difference because both of them used the same words.
This is fine for some questions and a disaster for others. The questions it is a disaster for are exactly the ones a person actually needs help with. What should I do about my father. Should I take this job. How do I talk to a teenager who has stopped talking to me. A language model will give you a confident, plausible, well-written answer to all of these, and the answer will be assembled from the average of every advice column ever published. That is not what you need.
What you need is the person who actually went through it.
“What you need is the person who actually went through it.” POST AI · On the library
What the library is.
The library is a record of things that actually happened to actual people, written down by them, kept together in one place. Each entry is small. Each entry has a question, a context, and an honest account of what was tried and what happened. My business partner stole from me in 2003 and these are the three things I wish I had done differently. My mother had dementia for eleven years and this is what I learned about being patient. I quit a good job to make pottery and it took six years to know if it was the right choice.
The entries come from the circles of this conference. They come from the main session in Izu. They come, in time, from anybody who wants to add one and is willing to write it down honestly. The library will not grow quickly. Real things rarely do. Over time it will become something that did not exist before: a searchable, browsable record of human experience from the people who actually had it.
The library will not be on a phone, or rather it will be on a phone but it will not behave like one. It will not optimise for engagement. It will not push you anything. It will not gamify your participation. It will sit there and wait until you have a real question, and then it will help you find people who have been near the answer.
Why it will work
Because the thing that has been missing is not information. We have all the information. The thing that has been missing is the slow patient transfer of what to actually do from people who learned it the hard way to people who are about to need it. That used to happen automatically because the older people lived in the same house and you saw them every day. It does not happen automatically any more. So we have to build it on purpose.
The library is the building. Each entry is a stone in it. Your circle puts in a few stones every time it meets. So does the Izu session. So does an elder who decides to sit down one Sunday and write one true thing she knows. The library does not care how famous you are or how articulate. It cares whether what you put in it is true.
How CCEF3 fits.
The library lives on CCEF3, which is the platform built originally for teachers. The reason it lives there is practical. The infrastructure already exists. Teachers are already using it to record what they teach and what their students learn. Families are using it through Randoseru to record what they choose to focus on each season. The library of the conference is the same idea extended to anybody who has lived a chapter worth writing down.
You do not have to be a teacher or a parent to upload to the library. You have to have something true to add.
Your own trail on CCEF3 is yours. Nobody else can edit it. Nobody can sell it. It cannot be transferred or inherited. It is a record of what you have contributed, which means it is a record of who you are choosing to become.
A credential. You cannot put it on a CV and expect a job from it.
An oracle. It will not tell you what to do. It will show you what other people did, and the consequences they reported.
Neutral. Every library has shelves and every shelf is a choice about what is worth keeping. We will make those choices honestly and tell you what they are.
It is not finished. It is barely started. Most of the books are unwritten and that is what your three hours of conversation is for.