# Brief 22 · Founder memo: the universe viewer, and where it goes next **date** 24 August 2026 · **kind** founder voice memo, transcribed by otter.ai, reproduced **verbatim** below including transcription artefacts. The instruction table and everything after the quote is the agent's reading, and is marked as such. --- ## The memo, verbatim > Dinis Cruz 0:00 > Okay, so this is the last memo to capture my sort of views on the the universe viewer that we have, which is starting to look pretty awesome, and I just feel like there's a couple little changes. Well, I think we need to keep improving what we're doing because I think this is so freaking cool. Okay, so first a couple little bugs. So when I open up the the tignogen graphs, the graph it doesn't fit. It's not in a fit state. So we should literally you should run the fit here, right? That basically shows you know the whole universe of the graphs that we have, at least the full selection that we have, right? A couple of little bugs is what I meant by the the box, and it's you know makes sense. It just maybe my brief wasn't clear. I'm not saying putting a box, rounded box around the label text underneath the circle, the the dot. I'm saying replace the shape instead of being a circle to be actually the box with a label inside of it because I think that makes it easier to read, and that means that when I'm reading stuff, I'm not. I don't have to find the dot, read down. Find the dot, read down. I basically just read the the bit straight away, right? And I click on it, and I basically go to each of the areas here, right? Which is pretty cool. Like the navigation is starting to work really, really nicely from one to the other. The again medium search large is great. The multiple views. Now, the other thing to add here is at the moment the doctory adds. So I think what we should have here is several several modes here, and and I I want you to start thinking in terms of sources of nodes to add. So it's almost like not necessarily a view, but sources that we adding to the thing, and this goes to one of the most important points I want to make in this voice memo, which is the the peaks and the centres of gravity. So, so what we have is at the moment we have the doctrine, right, which is freaking great, right? Exactly what I wanted is it's a way to navigate to see the document in a graph-like structure, which is freaking insane, and it works even nicely once we actually put in the. Oh, I just noticed one little bug. The when I click on the on the doc tree, I don't go to the section, so I should be navigating the website. And and and also, if you think about it, like the doc tree, what we have is exactly what we have on that navigation that we added to the source, right? So I should be able to click on it and go do actually what you do with the top navigation, right? That you added to the source document. So I should be able to click on it. So what we have here, especially on the tree, is you have this interesting concept of the top of the the pyramid, right, and it's almost like these what we call we call these peaks, right. So normally the peak is a document, right. So in the view, I need a way to see instead of being yeah the doctrine or document, and what I need is then you know when you have here each of these concepts on the left, where you have the dictionary, and then you have the the claims, and then you have the now the hypotheses, the ontology, you know the thesaurus, and and all of that, right? What's really cool, if you want to think about it, that each of those is a tree, right? Each of those is a tree, and we need to expand on that tree, which is coming next. But each of those needs to have a peak, right? Each of those needs to have its own peak, and the peak initially will just be a node that is hypothesis that we add to the graph, right, and what that means, it means that I should be able to, instead of the view, I should have like sources, and then I should have, and I should be able to select which sources I want to view. > > Dinis Cruz 4:10 > It's almost like which note packs we add in to the view, right, and this would allow me to see more or less. Let's say I just want to see the hypothesis, and I want to see the tree view, I can click on the doc tree and the hypothesis. Then I see both of them together, right? And this would actually clean up the view because at the moment, what you have is you have at the bottom of the tree you have all those bits that get linked, and I think there's a structure here, right? And that's the first one, right? And we, by the way, we need a global clean to button to just reset the whole UI because you start to do a lot of experiments here, right? So the other thing that now becomes super powerful is once you start to add these sort of doctrines, right? So these these bits with hypothesis, sorry, these peaks, is that if you think about that, the hypothesis should be again a tree in itself, right? So you should have the hypothesis. At the moment, we have a whole pile of them, and especially when you go to ontology, right? Each of those should have a tree, which again follows the idea of of of compression as you go up and expansion as you go down, but but in a way they should all be connected, right? Like in a way when I go meaning to connectivity, I should have notes that go down. Like where does this relate to? Right, so I should be able to see the relationships between the different concepts, and and what that means is that when I click on one of them, I should be able to now also navigate to the stuff that these means, and and here let's experiment maybe with stronger, weaker links. So maybe start with the stronger ones. Maybe have some more. Maybe have the option to have weaker if you feel like we need more. But the idea is to create a very rich view of hyperlinked nodes. So if you think about when I have a node and when I have a claim, or you know when I talk about my you know toss for example or my dictionary, I should be linking to the relationships of that right where you know what else we talk about. What else is related to this? If I talk about discover emitting or enrichment or graphs or graphs, what does this connect to, right? And and it doesn't connect to it, so that becomes quite interesting because what you then are creating is this visualisation of the terms, and it will be very interesting in a way. It's also become a nice deterministic way to find gaps or to find terms that are not well very well defined, where other terms, other concepts, they should have tonnes of connections. And yes, I'm very aware that this is going to, you know, explode our graph into that sort of dot and the big blob, which is okay because it just means that the models are very rich, and and then we we then find ways to visualise it and connect that. And one of the things that I would like you to experiment is then have the option on a config that we have at the top to, for example, see the paths to the peaks, right? Because at the moment it's really cool to connect, pick something, and then see all the paths up, all the paths down, and also maybe start to see those relationships, and maybe that one or two degrees of separation of those relationships. And remember, then one of the things that I think we capture here is that every one of these relationships should have a unique two-way conversation, right? One down and one up, and they should need to be unique terms so that we can again create the graphs from it, and we don't have the explosion of terms, right? One depends on the other, depends on the other, because because that's the interesting thing. Like when you have, you know, the claims or the Dinis or dictionaries, like there is a one-way relationship, sometimes with the or one way that one leads to the other, or one causes the other, right? > > Dinis Cruz 7:44 > So we need to make sure that it's not just related to, as we talk about in our document, right? So yeah, so let's experiment this, and and and that's why, for example, when you see the sub tree one, is this you're just seeing, you know, actually the sub tree one has a bug because that should only show the nodes that are selected, and then everything else gets frozen by the selections, and also yeah, in here you should also have that mode where I can add or remove like the freeze. It's almost like a freeze graph, right? And clean everything else, and I should be able to add or remove nodes maybe by selecting them, or maybe by seeing one node down. Maybe have the option to say, hey, show me one node down or two nodes down from here, maybe on this direction, a direction, and then actually when you do here, will be quite cool is to also show the the universe of the edges that are relation related here, right? And you kind of I think start to show this, for example, because you hear you got this thing where you say the fractal exhibits some graphs of graphs, but actually again, you saw the see look. You have fractal principle exhibit graphs or graphs. Like this, this should not be a two-way relationship. Should be a one-way where what does the is like fractist frac the fractal principle what graphs or graphs and the graphs or graphs what fractal principle the same way the graphs or graphs what recurs structure right which is actually I think is the the chapter that it's on right so again let's let's try to start visualising this and experiment this because I think we're really close to have something quite spectacular here. --- ## The instructions, and what each commits the work to *The agent's reading. Where a transcription garbled a term, the reading is noted: "tignogen graphs" is thinking-in-graphs; "doctrine/doctree" is the doc tree; "Dinis or dictionaries" is likely "the claims or the dictionary"; "two-way conversation" in context means each relationship carries a distinct verb in each direction.* | # | Instruction | What it commits the work to | |---|---|---| | 1 | The graph opens not fitted; it should open showing the whole selection | Run fit after the initial layout and after the panel first sizes, so the full universe is visible on arrival. | | 2 | Boxed labels: not a box under the dot; **replace the shape**: the node becomes the box with the label inside it | The boxed mode restyles nodes as label-sized rounded boxes, read directly and clicked directly, no dot-then-label eye movement. | | 3 | Clicking a doc-tree section node should navigate the source pane to that section, like the location trail does | A tap on a section node scrolls the source to that heading; the doc tree becomes the same navigation the trail provides, in graph form. | | 4 | Think in **sources of nodes**, not views: node packs added to or removed from the graph (the doc tree is one; each family is one) | The graph options gain a sources group: doc tree and each family are selectable packs; the view is their union. | | 5 | **Every family needs its own peak**: dictionary, claims, hypotheses each is a tree with a synthetic peak node added to the graph | Family peak nodes with containment edges to their members, so the tree layout shows multiple pyramids: the document's, and one per family. | | 6 | A global reset button: experiments pile up, one control returns the whole UI to its defaults | A reset control that clears the selection, filters, sizes and stored preferences for the document. | | 7 | Concepts should link to related concepts; experiment with **stronger and weaker links**, starting with the stronger | Asserted edges are the strong links; a derived, dashed layer (concepts co-claimed by the same claim) is the first weaker experiment, toggleable. | | 8 | The interlinked term view becomes a deterministic way to find gaps: an under-connected term is an under-defined term | Connectivity per term is visible in the graph; the blob risk is accepted, the finding is the point. | | 9 | A config option to see the **paths to the peaks** from a selection, up and down, and one or two degrees of relationships | A paths-to-peaks toggle: with a selection, the walks from it to every visible peak are highlighted. | | 10 | Every relationship carries a unique verb each way, never a generic relates-to; one term leads to or causes the other | Already the registry's law for asserted edges; derived edges are labelled as measured, and the display must read as directional. | | 11 | The subtree view "should only show the nodes that are selected, and then everything else gets frozen"; add a **freeze graph** mode: add or remove nodes, expand one or two steps in a chosen direction | A build-a-view workflow: freeze the current set, then grow or prune it interactively. **Needs the founder's confirmation of the semantics before building** (question 1 below). | | 12 | The fractal-principle / graphs-of-graphs edge reads wrongly: it should be one-way with the right verb, and graphs-of-graphs belongs to The Recursive Structure | The extraction's edge is reviewed with the founder (question 3): the data records `graphs-of-graphs exhibits fractal-principle`; if the reading is wrong the edge is corrected in the extraction, where it is anchored. | ## The questions the agent owes answers on 1. **The freeze mode.** Proposed semantics: a freeze control pins the current visible set; while frozen, tapping a node toggles it in or out of the set, and an expand control adds the neighbours of the selection one step at a time, with a direction choice (incoming, outgoing, both). Is that the intended workflow? 2. **The first weaker-link derivation.** Proposed: two concepts are weakly linked when one claim is about both (dashed, with the count). Alternatives: co-location in a section, or thesaurus adjacency. Which first? 3. **The exhibits edge.** The extraction records `graphs-of-graphs exhibits fractal-principle` (anchored to "At every level, the same principles apply"). Is the complaint about the direction of the fact, or about how the drawing renders direction? If the fact is wrong, the fix is one anchored edit in the extraction. 4. **The subtree meaning.** Today subtree shows the selection plus everything that hangs off it. The memo says "only the nodes that are selected". Should subtree become the frozen-set workflow of instruction 11, replacing the hang-off expansion? --- This document is released under the Creative Commons Attribution 4.0 International licence (CC BY 4.0).