# Founder memo: build the universe first, then find the plot

**Date:** 23 August 2026 · **Source:** voice memo, transcribed live by otter.ai and reproduced
verbatim below (per the house rule: the founder's voice is source material and is not edited)
· **Recorded while reading** [dev pack v0.3.27](../dev-pack/index.html), at section 5
· **Status:** it reshapes the pack. What it changes is listed after the transcript.

> **Where the earlier briefs are.** Briefs 00 to 19 belong to the first edition and froze with
> it at [v0.3.26](../admin/versions.html); they are at `/v1/briefs/`. This directory holds the
> briefs written after the freeze. The numbering continues rather than restarting, because the
> corpus is one sequence even though the editions are not.

**What it says, in one paragraph:** the pack is defining answers before the questions are known,
and section 2 is the clearest case: it hands every unit of the first edition a carry / lift /
rewrite / drop verdict before anyone knows what shape the second book is. The best thing in the
pack is the six node families, and the memo's central claim is that **nothing about the levels
can be known until that universe exists**. So the construction order inverts: not spine first
and evidence attached to it, but **all the reference material first**, decomposed into concepts,
facts, claims and evidence, and only then the plot lines drawn through it. A book is a journey
with pace, punchlines and characters, and the question the author is really answering is a
whodunit: what is the sequence, what is the story, what does the reader need in chapter one that
chapter two will spend. Three specific corrections follow: **fractal semantic graphs does not
mean one grammar everywhere**, it means semantic graphs nested in a fractal way, graphs of
graphs, ontologies of ontologies, with local override where a scope needs it, exactly as the
Issues-FS lexicon works; **Wardley maps are not an appendix**, because a map is a graph that has
gained position, movement and time, which is what makes a graph actionable, and a map of a map
of a map is the fractal claim in its most legible form; and **the five spine statements are not
yet the answer**, they are a guess to be arrived at rather than assumed. The frame for all of it
is explorer to villager and back to explorer: the first pass ran to something near a product,
it has been frozen, and this pass starts again with better tools rather than from scratch.

---

## The memo

> So, okay. So, me qualify as a voice member, but I actually think that you should create first
> a new Markdown document from this, and I'm gonna ask you to do it here, not on the other one,
> because you have a better context of what we talk about. So, you should be able to make more
> sense of the references that I'm providing here, but so my thinking here is, by the way, the
> brief is amazing, right? Like it's really cool. I'm really good. I'm I'm on I'm on section
> five, but one of the things I'm already noticing is that we are already defining the answer
> before we know what questions we're answering, and and this is very obvious in section two,
> where you say what exists today, because this is already making a number of judgement calls,
> but we don't know what the shape of the book is, and even when we talk about the the shape of
> the book, where I think is very interesting, and I think the most amazing things we've done
> when we mapped the book and we created the concepts maps is this section where you talk about
> the the book's graphs hold six families of nodes, and we can add another. But I think this is
> these are very good, which is the unit, the concept, the fact, a search, an opinion,
> hypothesis, an objective. And I feel, if you think about it, that we cannot know what each of
> the levels, level one, level two, level three, will work will look like until we have done
> this. So, so I think this is like a whodunit, right? Like we need to come up with the plot. We
> need to come up with what is the sort of the whodunit mystery machine. Like what is the
> sequence of events? What is the narrative? What is the story that we want to tell on the book?
> What is the journey that we want to bring in? So, so I feel the way to do this is actually a
> bottom-up approach, and I think you already have most of what we need. But I think the logic
> here, and what I think we should be capturing in the brief, is that this is why it's a brief
> to create a brief. Is that we don't know which parts of the first version of the book we will
> take. We, well, right now we have an idea, but the exercise that needs to be done is an
> exercise where we need to take all of the reference materials. We need to take for all of
> them. I think you already have done most of it. What are the key concepts, the key facts, the
> key researches that we're trying to make, and how does the evidence look like? What does that
> graph looks like? And then we need to create those plot lines. We need to create the plot
> lines in a way to almost to think about how to pace the user. You know, it's the whole thing
> of how do we introduce the concepts on chapter one that's going to be needed on chapter two?
> How do we guide the user, and also more importantly, once we start getting feedback from the
> users, how do we make sure that, in a way, the readers can help us to say, oh, I didn't
> understand this concept, which became clear later on, or oh, this was a great way to introduce
> the storyline, and if you think about it. We we we already have a really good baseline of
> data, right? And we have a really good baseline of of concepts. In fact, even the way this
> book is being created, it's a great baseline of showing the graph because all this is a graph.
> Like we treating the creation of the book as this graph. So so in a way, although level one,
> level two, level three, level four is becoming the projection, and I think you correctly
> picked up the whole thing of when you go from level three to level two, you are compressing,
> you're gaining in terms of conciseness, but you're losing in terms of detail. And then when we
> go to level one, we we have the top level, so we it's almost like, I guess my instinct is that
> it's probably a level three, is where the action will be, because level three is probably the
> most the level that will have the most connection to the the graphs that we generate, and then
> level four is like you said, is the expansion into the multiple sections, and then level five
> is the prose, which is the book, and level two becomes the compression, and level one becomes
> the super summary, right? The one or two paragraphs where everything is linked down, and you
> know it's like the whole thing, right? In a way, if you think about on the level one, like
> you've done already, like every section, every paragraph, every almost word or two words links
> to a whole different thing on level two or even level three or level four, right? So the user
> can see it from those levels, but but we we need to know the plot. So this is like a murder
> mystery, right? Who done it? We need to figure out, or like a an adventure, like we need to
> figure out the journey because that's what a book is. Right, a book carries the user through a
> journey, tells a story, and it's important to have a pace. It's important to have punchlines,
> and I think we have a lot of characters. We have a lot of great nuggets that should make the
> reader think, which is what we want, and and in fact, some of these you can even say the
> objective might be what we want the reader the reader to agree once they see this, right? So,
> so that's what I think that the first thing that we should be building in this version two is
> this universe. Basically, it's this graph version of the book, where I as the author, can
> basically kind of go from my head, what are the key messages that I want to say, and then for
> each of those messages, we then link the evidence to show the stuff, the points, the actions,
> etc. And then I can see the book from multiple angles, so I can basically see the different
> peaks and the different valleys, and how how they go, and take it to account that sometimes
> you might have some concepts that then funnel up, right? There might be you know transversal
> concepts throughout the whole book that might be that those peaks that might not be on level
> one explicit, but they or they might be, or not, but they they end up in the the book
> throughout, which is fundamentally the instinct that I have, right, as an author. So I give
> you a good example. Like the reason why worldly maps are very important is because worldly
> maps are a great way to explain a lot of the evolution, and it's a great way to visualise
> graphs, and it's a great way that you can get meaning through connectivity because you also
> have the the fractal element of a map of a map of a map, and we have the you know basically
> because a map is a graph with basically more some interesting properties like position matters
> and movement matters, and also the the maps, especially when you look at different types of
> evolution and different types of of of of like horizontal, because it doesn't need to be
> generous to custom build. It can be different types of evolutions, right? It's a great way to
> make the graphs real and actionable, and I think that's very important that we have here
> because, you know, those feedback loops and those, you know, for example, one one of the
> things that for me is very important is the feedback loop that makes this work. It's about the
> economic models that work. So, for example, at the moment, so think about it. I need a way
> that I can say when I'm looking at the graph, will to say, all right, cool. Let's look at the
> economic arguments. Let's look at the feedback loop arguments. Let's look at are we capturing
> the importance of creating a system that will make a graph, for example, to be self-maintained,
> even when the people playing the graph don't care about it, or don't don't want to do it, or
> is even against their own interest to do that. And I think that's that's very important to
> capture that. And and one of the interesting concepts here, where when I was reading this, the
> the some of the concepts, and and and actually I think is the argument. Let me just find it
> here. Is for example, when you said in the argument section, for example, you have the spine,
> which of course, like, see, you're already making, or we already making a lot of decisions
> here. That you know, we, it might be the top, it might not be. So you got fractal semantic
> graphs, meaning to connectivity. The graph is the author's and the gimme is the finding.
> Confidence is a function of connectivity. So you already got these five key elements, which
> might not be level one, right? Like you know we we we should not assume that this is it. But
> also when when you look at fractal semantic graphs, you talk about the same grammar ever
> altitude one type system. See, I actually don't agree that this is what I want to say with
> fractal semantic graphs. The problem is I want to arrive at this. So I think the way I'm
> looking at this is that by the time I read these spine five statements, the the vision that I
> have for the book's concept and story should match what's written, right? So the reason why
> this is clearly not correct, the way I look at it, is because for me, the whole concept of
> fractal semantic graphs is literally semantic graphs in a fractal way, right? Where you have
> graphs of graphs or graphs, ontologies, ontologies. It's like what we did with issues FS,
> where we are overriding. We allow, you know, a lexicon to be created, but we allow that dynamic
> flow of each group overriding when needed specific things. That's what I mean by by the
> graphs. But again, we should be evolving into it. So, all right. So I think I think it's worth
> creating a memoir of this. Let me know what you think. But because the concept here is what we
> are figuring, and by the way, we're figuring this out. We might need a couple more versions to
> do this, right? And again, the beauty now, again on the worldly maps, is that this is where we
> go from explorer to villager, back to explorer mode. But every time we do it, we're getting
> better. So if you think about like we did our first Explorer session where we went from
> Genesis custom build, and by the time we were getting almost to a product, we go back to the
> Explorer. But now we have better tools. We refactor it. We did a move. We freeze what we had,
> and now we go back into the Explorer. But now we are evolving, so we don't start from scratch,
> right? Again, good example of worldly maps. Why they are critical of this, right? And why
> also, you know, the way I look at is that a map is an evolution of a graph, but you can't have
> maps without graphs. But maps is when graphs take a whole different meaning and positioning
> and strategic and gameplay that you can't just have with a graph. You know, graphs, maps have
> movement. Maps have strategies. Have time. You can reflect time in a map that you you really
> struggle in a graph, right? So, cool. All right. Hopefully, this makes sense.

*Transcribed by otter.ai. Reproduced without edit, including transcription artefacts: "a search"
is assertion, "worldly maps" is Wardley maps, "the gimme is the finding" is the divergence is the
finding, "genesis custom build" is genesis to custom-built on the Wardley evolution axis.*

---

## What each instruction commits us to

| # | The instruction | What it commits us to |
|---|---|---|
| 1 | "we are already defining the answer before we know what questions we're answering ... very obvious in section two" | The audit's verdicts stop being decisions. They become **hypotheses to be tested against the universe**, and the file says so. |
| 2 | "the most amazing things we've done ... is this section where you talk about the book's graphs hold six families of nodes, and we can add another" | ADR-3 is promoted from an architecture decision to **the first thing built**. And the families are open: a seventh is expected. |
| 3 | "we cannot know what each of the levels will look like until we have done this" | The phase order inverts. The universe precedes the spine, and the spine precedes the levels. |
| 4 | "this is like a whodunit ... what is the sequence of events, what is the narrative, what is the journey" | **Plot lines are a first-class artefact**, not an emergent property of chapter order. A book is a journey with pace and punchlines. |
| 5 | "the way to do this is actually a bottom-up approach, and I think you already have most of what we need" | Construction is bottom-up from the twenty-one carried sources, the vault analyses, the concept map and the first edition. **Reading stays top-down.** The two are not in conflict and the pack must say so. |
| 6 | "this is why it's a brief to create a brief ... we don't know which parts of the first version we will take" | The pack's own status is corrected: it plans the work that decides, it does not decide. |
| 7 | "how do we introduce the concepts on chapter one that's going to be needed on chapter two" | **Pacing becomes checkable.** See the proposal below: a concept used before it is introduced is a computable defect. |
| 8 | "once we start getting feedback from the users ... I didn't understand this concept, which became clear later on" | Reader feedback attaches to concepts and to the order they arrive in, not only to pages. |
| 9 | "level three is where the action will be ... level five is the prose, which is the book" | The altitudes are re-described: L1 super summary, L2 compression, **L3 the pivot with the most graph connection**, L4 the expansion, L5 the prose. This is a change from the pack, which had L5 as a reference layer. |
| 10 | "sometimes you might have some concepts that then funnel up ... transversal concepts throughout the whole book that might not be on level one explicit" | A concept can be load-bearing without being visible at the top. The peak formula must not assume that importance and altitude are the same thing. |
| 11 | "fractal semantic graphs is literally semantic graphs in a fractal way ... graphs of graphs of graphs, ontologies of ontologies ... like what we did with Issues-FS, where we allow a lexicon to be created but allow that dynamic flow of each group overriding when needed" | **The definition in the pack is wrong and is replaced.** See below. |
| 12 | "Wardley maps are very important ... a map is a graph with position and movement ... a map is an evolution of a graph ... maps have time, which you struggle to reflect in a graph" | **The audit's verdict on chapter 11 is reversed.** Maps are candidate spine material, not an appendix. |
| 13 | "the feedback loop that makes this work ... it's about the economic models ... a system that will make a graph self-maintained even when the people playing the graph don't care about it, or it is against their own interest" | A named thread that the current material barely carries. It is a hypothesis, and it is testable. |
| 14 | "these five key elements, which might not be level one ... we should not assume that this is it" | The spine is a **candidate**. It is arrived at, not declared. |
| 15 | "by the time I read these spine five statements, the vision I have for the book's concept and story should match what's written" | The acceptance test for the spine is the author's recognition, and that is a legitimate test to state. |
| 16 | "explorer to villager, back to explorer ... every time we do it, we're getting better ... we don't start from scratch" | The frame for the whole exercise, and the reason the freeze was worth doing. |

---

## The correction to *fractal semantic graphs*

The pack defined it as **uniformity**: one grammar, one validator, one query engine at every
altitude. The memo says that is not the claim.

The claim is **composition with local override**:

> Semantic graphs nested in a fractal way. Graphs of graphs of graphs. Ontologies of ontologies.
> A lexicon exists and is well connected, and any scope may extend, specialise or override it
> where it needs to, dynamically, without asking permission of the centre.

These are genuinely different claims. Uniformity is a constraint that makes things comparable.
Composition with override is a **capability** that makes local meaning survive without
fragmenting the whole. The second is the interesting one, and it is the one the corpus already
argues: *Issues-FS Lexicon*, carried at `/v1/docs/issues-fs-lexicon.html`, says in its own words
that the Lexicon is **not** a schema registry and **not** the authoritative source of
definitions, and that any scope can extend, specialise or override a bootstrap definition. The
carried source contradicted the pack's definition and nobody noticed, which is a finding about
the pack rather than about the source.

Uniformity has not disappeared. It is a consequence: what makes the override safe is that the
grammar for expressing an override is the same at every level. But it is the mechanism, not the
claim, and the pack had them the wrong way round.

---

## The correction on Wardley maps

The audit dropped chapter 11 from the spine on the grounds that nothing in the concept map
depends on it. The memo gives the reason that was missing, and it is a strong one:

- A **map is a graph that has gained position, movement and time.** Position is meaningful,
  movement is the point, and time is expressible in a way it is not in a plain graph.
- A **map of a map of a map** is the fractal claim in its most legible form, which makes maps a
  demonstration of the title rather than a digression from it.
- The evolution axis is **not** fixed to genesis-to-commodity. Other evolutions exist and can be
  mapped, which generalises the technique.
- Maps are what make a graph **actionable**: strategy and gameplay are things you do to a map.

The honest summary of the relationship, in the founder's own framing: *you cannot have maps
without graphs, but maps are where graphs take on meaning, positioning and gameplay that a graph
alone does not have.* That is a claim about the whole book's subject, and it belongs in the
universe as a candidate spine statement rather than in an appendix queue.

---

## Two proposals this memo makes possible

Neither is in the memo. Both follow from it and both are computable, which is the test this
project applies to its own ideas.

### The pacing gate

Instruction 7 asks how a concept needed in chapter two gets introduced in chapter one. Once the
universe exists, that is arithmetic rather than editorial instinct:

```
For every concept C and every unit U that uses C:
    introduced(C) = the first unit, in reading order, that defines C
    FAIL if position(U) < position(introduced(C))
```

A concept used before it is introduced is a pacing defect the build can name, per level, with the
exact pair of units. It does not tell you the writing is good. It tells you the reader was asked
to spend something they had not been given, which is the specific failure the memo describes.

The same structure serves instruction 8: reader feedback of the form *I did not understand this,
it became clear later* is a report that `introduced(C)` is later than it should be, and it
attaches to the concept and the pair of units rather than to a page number.

### A stopping rule for the universe

Building the universe first has one obvious failure mode: it never ends. A stopping rule is
needed before phase 1 starts, and it should be computable rather than a feeling. Proposed:

> **The universe is sufficient when every candidate plot line can be traced end to end through
> it without a gap**, and every spine candidate has converging evidence or is explicitly labelled
> an opinion.

That makes "enough" a property of the plot lines rather than of the material, which is the right
way round: the universe is not trying to be complete, it is trying to be **sufficient for the
story we are choosing to tell.**

---

## What changes in the dev pack

| File | Change |
|---|---|
| `00__README.md` | A fourth governing rule: **the universe precedes the plot, and the plot precedes the levels.** The status line records that the pack is being revised by this memo. |
| `02__what-exists-today.md` | Every verdict is re-labelled a **hypothesis**, with the memo's reasoning stated at the top: the shape of the book is not yet known, so a verdict is a guess with a reason attached. The Wardley verdict is reversed outright. |
| `04__the-argument.md` | The spine becomes **candidate spine statements**, explicitly not level 1. The fractal definition is replaced. Wardley maps move from "not on the spine" to a candidate. The altitudes are re-described per instruction 9. |
| `05__architecture.md` | ADR-3 promoted to the first thing built. A new ADR on plot lines as a first-class artefact. A new ADR on the pacing gate. |
| `08__implementation-plan.md` | The phase order inverts: **phase 2 becomes the universe**, built bottom-up from the carried sources; the spine moves to phase 3 and is derived rather than authored; the levels follow. |
| `09__verification-and-open-questions.md` | The stopping rule becomes an acceptance criterion. Question 4 (Wardley maps) is answered by this memo and closes. |

---

This document is released under the Creative Commons Attribution 4.0 International licence (CC BY 4.0).