# Founder review: a book for humans and agents

**Date:** 22 August 2026 · **Source:** voice memo, transcribed by Otter.ai, reproduced
verbatim below (per the house rule: the founder's voice is source material and is not
edited) · **Status:** Review **r002** on [the reviews register](../reviews/index.html),
normalised into nine items with the agent's comments and proposals at
[/reviews/r002.html](../reviews/r002.html). No content changes were made from this memo
directly: per the review workflow, the agent comments and proposes first.

**What it says, in one paragraph:** embrace what the book already half-is: a book for
humans **and** agents, stated in the identity and given a proper introduction. The
agent guidance should read like briefing the agent that is helping write it: name where
this material deviates from the standard usage already in an agent's memory (property
graphs, live graph databases versus the ephemeral serverless pattern here, acceptance,
the fractal concepts), and carry indexes for follow-up. The book itself is a graph and
should exist as one: each section with its facts, hypotheses, its own ontology and
defined terms; a machine-readable graph version, more expanded than print, with an
abstraction layer that can show just the facts, the key concepts, the narrative and the
objectives of a chapter. Words alone carry less meaning than words backed by evidence,
so the book needs a references and evidence layer (JSON files, cross-indexes, sources):
the 10,000-hours story is mentioned but not sourced, and "how do we know our version is
correct?" is exactly the question. The designed use is that a human hands the book to
an agent and asks "check the references, how much of this is correct?", and summaries
and infographics are made on demand: dogfood that on this book. Each chapter should get
a graph visualisation showing how deep the evidence goes, including assertions with no
evidence connected yet, shown honestly. Provide the machine surface as tools agents can
use. For the printed version, links must become footnotes, because hyperlinks die on
paper. And the missing pieces of the puzzle belong in a separate mapping, which is
where the graph of evidence starts to be visible.

---

## Verbatim transcript

Dinis Cruz 0:01

So I want to expand on another concept that on a book that we started with, but I think
actually there's something here, and I feel like we should just embrace it, which is
this idea that, and even the title should probably mention this, and we should have a
good intro for it. That this is a book for humans and agents, and I think that's
important because I think what's interesting about this is on the agent's guidance,
it's almost like you know this is Claude that you're helping writing this. You know,
it's a bit like think of the bits of the content that are not normally in your normal
memory and the normal usage of terms. So you know, for example, you know, like that's
where you have those references to property graphs, and have a reference to how people
use graphs and how usually use graphs database and how they don't. have the sort of
ephemeral concepts that we are have here, and so there's a lot of interesting concepts.
You know, when we talk about acceptance and fractal stuff, there's a lot of interesting
concepts here that the the You know we should be doing the brief, and that's basically
what the brief. You know I like those sections you have for agents, and also what you
should also be doing is creating the materials for the agents, right? So for example,
you know especially when you have all those links and all that stuff, I think that's
more for agents than for humans, Because ultimately, what's interesting about this is
that if we think who would correct, yeah, if you think about it, actually, the links
that we include are for agents, and actually here's the angle for this. In fact,
several angles to take here. First of all, the book itself is a graph, right? And we
need to apply the same logic that we had on the other concepts that we have, where each
section needs to have facts, hypotheses. It needs to have its own ontology, right?
Needs to have their own terms that we define, because again, we get meaning to
connectivity, right? So it's the same thing as, if you think about it, we should have a
graph version of the book, because that's ultimately how we can also represent the
multiple elements, right? And be navigated And if you think about it, the idea here is
that we should have a whole section of references, JSON files, cross indexes, you know,
sources, links, etc. which fundamentally provide the body of evidence and is the
follow-up. In fact, for example, the analogy that we have of what's it called? The
100,000 hours, 10,000 hours. Again, we we mention, but we also don't source it. So how
do we know that our version is the correct one? So so it's almost like if you think
about it, that is a fact that they need to be verified. So I think that in a way, and
and there's so this also means that the web version of the book, and the graph version
of the book would actually be a lot more expanded, where we don't have to represent
everything in a book, right? And so, so we need to. So I guess one of the things that
is interesting to find ways to represent in the book is what are those cross
references, so that when the agent is reading this, the agent can follow up. And I
think we should explain this. So, should explain this that the expectation here is not
that the human is going to follow up the clues. The expectation here is if the human is
reading this, or the human can basically give the book to an agent and then ask the
agent how much of this is correct, how much of this is, you know, if I'm going to trust
this book, can you check that the references are correct, right? Can you check that
this is accurate? And again, we should be making the book so that if somebody just
wants to get a summary, they should be able to create a summary. Someone can get an
infographic, create an infographic, right? So we should be leveraging in a way the
technology that we have available to to share this thing here, right? Because
ultimately, that's part of the the workflow that we're showing. So I guess what's
interesting about here is this idea that this is a book for humans and agents, and
actually the website and the the graph almost structure the graphs and the semantic
graphs and the fractal graphs that we will provide on our website, which will, I guess,
will be interesting to see is if we can then, for for each each chapter, we can provide
a graphical representation of a visualisation of what that graph means, almost to say,
hey, here's how deep the agent can go. Here's how deep these evidence is, and that's
okay because you can have a series of events where I'm making assertion that there's no
evidence, or we haven't connected the evidence so far. Like we said, that means there's
less meaning on the words because, after all, you know, the words of the book are just
words, right? They should be backed by evidence, so they should be linked to other
concepts. And in a way, my I should be able to hyperlink to the stuff that I publish,
but then I should be trying to find evidence for others who say the same thing, right?
So then that builds the body of evidence for the topics that I'm talking about, and I
think that's an important concept to explore here. And and again, when you say it's a
book for humans and agents, means there's also a part of the website that is very
machine readable and agent driven. Even have the tools that we can provide, so we can
even say we provide a set of tools that the agents can already use to make their life
easier, right? So again, I like the idea of somebody can take the book, and I should be
able, even I should be able to go to a new agent to say, hey, can you check the
references on this book. Can you do a research on this book and check if the references
match up, if they are correct, if they are accurate, and how does it goes from there?
Right. So, again, let's let's dog food the book themselves, right? And and that should
make the book stronger because if you think about it, then in a way the book is a
graph, right? And that's how I should also be managing because this is the whole
concept, right? If you think about it, the the book should be a set of graphs and
hypotheses, and in fact, we should have an abstraction layer of the book where I should
be able to just see a version of the book that is just facts, that is just key
concepts, that basically shows the narrative and shows the objectives and shows what
what are the key points that we're trying to make in a particular chapter, and what are
we trying to show? So so again, that guidance is focused. So so for wrapping up the
thing, when when we have these comments also for agents, it's important to say how to
think about the page and where does this deviates a little bit from standard
literature, or maybe have that's where you can then put indexes for it for the agent to
go and find more stuff. The other thing that I feel I didn't mention the previous one I
should mention is that, especially for the printed version, I think especially links
between things, I think they should be done using footnotes because the hyperlink it
works on the web, but doesn't work. Actually, works on the PDF, but won't work on a
printed version. So we kind of need to think of the printed version that has there, and
and also there's a bunch of comments in there about some of the missing piece of the
puzzle, right? Again, you need to think of this from a book point of view. We should
have all those missing pieces on a separate sort of mapping. And actually, the way
we're describing right now, we start to see the graph of evidence and how everything
fits together. So yes, I think there's some cool ideas on on this memo. Again, let's
capture this again on a review. So this is another review for you to capture to create
a debrief and then to map out how would you change the book from here.

Transcribed by https://otter.ai