 The T2 Tile project is building an indefinitely scalable computational stack. Follow our progress here on T Tuesday Updates. Welcome to the T Tuesday Updates, our top stories this week. Scaled corruption. Yeah, there's a guy, Ape Panwar, I guess his name is Avery Penrum, who has had a blog that puts very thoughtful, you know, good nerd, good engineering solid stuff that I've been following for months or years, I guess, had one out recently that I wanted to talk about briefly. Absolute scale corrupts absolutely. Now, since I talk about scaling and indefinite scalability, if we're saying that the scale is bad, I need to know about it. But in fact, when we drill into what he's got to say, you know, so the idea basically is that, you know, we used to think the internet was going to just make everything good because it's going to make transparency and interconnectivity and people will talk and when they talk they won't fight. That turned out not to be the case. And what he observed is that across many different kinds of systems, not just the internet, when anything scales up, it gets corrupted, it gets invaded, it gets exploited from the internet to banking systems to space debris, everything. And it has a sort of biological aspect to it and therefore one of the defenses that he's talking about is diversity, which, you know, as having had some academic experience doing biological approaches to computer security, you know, yeah, we pushed diversity. And that seems pretty good. But really when you look more deeply into it, it's not, to my eye, it's not actually the scaling itself that he's objecting to, but the manner in which the scaling is done. It's about the, you know, the centralized stuff, to say scaling means you have one thing in the center, you have Facebook, you have Twitter, whatever it is, and then you have this flat structure and that's what scaling means is, well, yeah, that's not going to work. I mean, that's CPU and RAM writ large, you know, and in my analogy, you're going to see exactly the complete inability to secure the system once it gets big. And the way that you need to scale is in a sort of communities of communities of communities, communities all the way down, where you actually have to stop and figure out what the appropriate level of structure is at each scaling interval rather than thinking you can solve it once and then go flat. I don't know how to fix giant social networks, but I do know a general workaround, slow things down. Yes, not as efficient as possible, as efficient as necessary. So this is very much in the spirit of robust first and best effort and all of the stuff that we're working on, the Feast Machine and the TDutile project. So I wanted to just call it out a little bit and highlight it. Most interactions should not be internet scale. Yeah, yes. And APenwar is now involved in some company called TailScale, I think, which I'm not sure exactly what they're doing, but just from a, from a miles away, it looks like it's kind of, I'm guessing it's a sort of crypto overlay for, you know, business logic on top of the internet, something like that. I'll have to dig into it further and find out, or if anybody knows, you can tell me in the comments whether I guessed right or not. Most interactions should not be internet scale, yes. But what we're suggesting here is that we want to reintroduce explicit space into the scaling process. So you'd say, yes, in fact, this interaction should be one meter scale close to my body. This interaction should be one kilometer scale, something like that, and use the properties of physical space to actually get some relatively much stronger and not arbitrary guarantees about who's in and who's out as far as the interactions that they can take place. Anyway, so absolute scale doesn't corrupt absolutely unless you do it in the centralized flat thinking that it's all going to work as opposed to needing to rework it as you go. And that's what we're trying to do with the T2 tile project. We're trying to rework it bottom up. All right, the power zone frames started about it last week. We now have a full box of them. That's seven power zones. We need something like 10 or 12. So we're more than halfway. If things go well, chances are the power zones will be done next week. Yeah, so last week we were working on bonds, trying to build a standard library for bonds, something like that, to make it easier to write elements that have relationships to other elements that even if they change locally in spatial orientation, they can still remain connected in a logical fashion. We had bugs last week. We've now got it working. One of the good things, so the way this whole structure works is we're doing bonds by multiple inheritance. So you inherit from this kind of bond. If you have that one, you inherit from whatever kinds of bonds you have. That means that the base classes that are doing the general logic don't specifically know what bonds you really have. So they call down in virtual methods just to let you, the guy at the bottom, please check the consistency by calling back up for each bond that you've got and so forth. And this is working fairly well. It requires a certain amount of boilerplate that there will be nice if we get rid of, but really it's fine. So we had our bondos last week that we could make a seed of them, make a seed, and then it would pop out into a head, middle, and tail. And now they work fine. They rattle around great. And they can cover quite a bit of distance. And again, they still maintain their logical relationship because they have a bi-directional, just a quick reminder so the tail bondo has one to someone ahead of him. The mid bondo has one someone before and someone after. And the head bondo has someone just behind him and so forth. And this is all being done in nicely, reasonably structured, and templated form. I took it further than that. And so the seed bondo idea, the idea of always beginning with a seed and having it pop out, the structure that you want to sort of create the garden of Eden structure that you need for a particular computation, that's the halfway mark. It's not like serial determinism in traditional CPU and RAM computing where you don't use everything absolutely under control. And you can have a giant, completely complex structure that you then unfold. But there is an even more sloppy bottom-up way to go, which is instead of having a seed, just having everything opportunistically generate themselves as they go. So I created these M2s, which are meant to be just simple, pairwise. Each has two bonds that it can connect to. And if it finds, it just looks around in the world, so we can make one. All right, so there's a whole pile of them. And initially, they're all un-bonded. They have the capabilities for two possible bonds, of which they've got initially none set. But they just very quickly now start to polymerize as they randomly look around at their neighborhood and find, hey, I have a head bond available. He's got a tail bond available. Let's join them up. And so I added an index there. You might see it flashing in and out in the Adam View panel so that basically each guy tries to be one more than the guy downstream of him. And once the monomers have formed a loop, which they don't have to form, but they typically will. And they'll often form dimers. Just one guy goes here, the other one goes right back. And these structures will also have quite long polymers that'll pop out of these things just at randomly. They don't make these big, long loops. That'll try to count up each one trying to be one more than the less. And when it does create a loop, it'll just keep going until it hits the maximum value that the number can represent, the count to infinity problem of distributed systems, in which case what this does is it just resets itself back to zero and then starts counting from there. And that is now expressed in the color. So when it's reaching the maximum value, that's full red. When it's the minimum value, it's full green. So you can see when there's a loop, you can see this red, green translation going around. It's actually kind of fun to watch. I've done some time lapses of it, but we'll let that go for now. All right, so we have bonding. And that's the state of the artificial chemistry. Oh, and there is also some developments on intertile events. We're trying to build up the state machines that are needed to coordinate between two tiles when they first interact with each other. And also when they're trying to race with each other about who can have an event on overlapping corners of the world, getting the state machines right was not being all that easy. So in fact, this week I developed a simulator of the simulator just for purposes of studying the state machines. And I thought I would show it to you real quick. It's a curses-based thing. So this is actually an entire power zone. That's four cross, four rows, four columns, and so forth. Right now, all the tiles are off. So the first step is turning it to passive, meaning it'd be willing to accept events that other ones are doing, but it won't originate events itself. At that point, the six intertile connectors all get enabled, but there's really nothing more they can do because the neighbors are turned off. But if we turn on a neighbor, they're both there, and now East and West have discovered they're in contact and so forth. And we can take, we can issue commands to the whole fleet here like that, and we can turn things down and watch stuff adjust. So the state machines still need work. We got two state machines, one for each intertile connector and one for each tile as a whole that are then duplicated over all these guys. But this makes me feel a lot more confident that we're gonna be able to make a state machine that makes a fair bit of sense and feel good about what it's trying to do. Star minus, star minus, turn everybody off. Okay, so that is the progress in the intertile events zone. And finally, the month of November. So one of my kind of bucket list things was to write a novel or more specifically, finish a novel. I mean, I've never even really gotten very far into I've written short stories and stuff over the years of course I've written lots of academic papers and so forth but to actually do a full on chapter after chapter, big story, lots of plot and so forth, I've always wanted to do it and there's this thing. Do you know about this? Nano Remo, Nano Remo is the national novel writing month and it's November and I'm gonna try and do it. I'm gonna take the month of November 2019 right here in the middle of the t2tile project and stop and try to write a 50,000 word novel. Now 50,000 words it turns out is not even that big a novel but that's 1600 plus words a day every day in the month of November. Am I gonna really succeed at this? I don't know, I really don't know but I've been thinking about it for the last several weeks I have a sort of a plan, a plot, a story, a framework that actually connects to the t2tile project. So this week I signed up at nanoremo.org. I'm Vaughan Joy Manin, the author of The Path, The Best Effort and I started the project and the name of the project is Best Effort. It's a science fiction novel. It's set 20 years in the future of The Path, The Best Effort so that means it's around 2050, late 2050s, something like that, not exactly short of a year. And it's a classic story of a boy on his computer. Sort of, we'll see. I don't know, should I be still trying to come here for tTuesday updates and say, okay, now I've got, the goal is 50,000 words, 50,000 words. I really don't wanna let myself get pushed off in thinking that I can go hack and call that productivity for the week. Some of these 50,000 words are gonna be garbage but this is part of my own thing. I tend to be a perfectionist and as a result I write very slowly. The part of this is to get out of that and to get this stuff out. And certainly the whole nanoremo spirit is that you're allowed to expect to revise later and this is more about getting a completed draft than actually polishing something that could be ready to get representation out in the world of fiction. We shall see, but it's got a little bit of a justification. I mean, because a lot of times people want me to be more cutting loose about the vision about what could the T2 tile project, what could the movable feast machine, what could best effort hardware really do? How could it really work in the future? And I always resist it because I'm trying to stick to the science. And so here's a way to cut loose. We'll see. I don't know. I'm super terrified. I made this a public project at nanoremo.org. If anybody wants to stop by and give me support or however that works. I don't even know how that works yet. I would be eminently grateful. And once again, if you see me around too much it's you can't go wrong by saying, Dave, why aren't you writing? So that's it. So will the next T Tuesday update be out next week at noon mountain time on the November 5th? I don't know. I don't know. What do you think? Thanks for being here so far. Have a good week. If I don't see you, have a good month in November.