 Hey folks, it's T Tuesday. Welcome. Thanks for dropping in. I made a new friend. Take a look. Meet BV, the full self-driving vehicle. BV falls from the sky and drives around on a checkerboard. Now, you may be asking, where's the driver? Well, BV's the driver. Wouldn't be a full self-driving vehicle, would it? BV's got sensors for the color yellow to the left and above and to the right. And BV's really into the color yellow. But BV has a problem. Because unless BV ends up all alone on the checkerboard, which happens sometimes, what otherwise happens is this. It just keeps happening and happening and happening. New tires. Big motors didn't help either. Need some impressive moves. But the final result was the same. BV had a dream where front spoilers in a nitrous kit helped catch the yellow. But then the Great Spirit VB appeared and explained that the real problem was bad wiring inside. More yellow caused more velocity. What BV thought was love was more like aggression. But yellow is so exciting, protested BV. If more yellow caused less velocity, things could be very different. But sorry, BV, you're just not built that way. I wish to change my wiring, pleaded BV as the Great Spirit began to fade. And when BV awoke, his wiring had changed. Sometimes he still ended up in the void. But most of the time, he lived happily ever after. So that's BV. You know, it's funny. I sort of felt for a long time that I wanted to make some kind of anthropomorphic thing to go with the T2 tile project. But it didn't seem right to try to make something that some animal that represented the tiles all connected together or something like that. But now that there's a body, which is, you know, more or less connected to the things, I'll explain what level of connection we have for that in a second. But then it was okay. It was all right to make up a character and play along with it. So I sort of had a little fun. All right. So this was the goals for today. Breitenberg car controlled by the T2 matrix brain almost got that. The signals go all the way to the T2 tile, but not actually to the matrix yet and have gigantic fun. Yeah, we did pretty good this time. So okay, so we met BV. I want to review the schedule for a little bit, then talk about the sensory motor stack because it's a huge pile of stuff. And there were some interesting issues in the in the engineering of it, you know, other than just, you know, my normal stupidity and well, and some some some good design stuff, I think. And then I want to take a minute at the end to make sure we have time to talk about community stuff because I haven't done a while. Okay, so here's the official spreadsheet from all the way back in the beginning. There's all kinds of complex stuff here. But for me, the bottom line is we're halfway done. We got three episodes before three episodes after here we are in the middle. You know, and the thing that's amazing to me is, you know, I haven't written any Oolong code in months. Yeah, I don't think there's been a time that I haven't written any Oolong code for this long in in years, maybe closer to a decade even, I don't know, which so it's been good and bad, you know, I'm writing a lot of Python. So the goal for today was to demo Breitenberg vehicles. Well, we did that. That's what BV is, a Breitenberg vehicle. I'm going to change the schedule a little bit going forward. Instead of expanding the matrix beyond 100 tiles, because honestly, at the level of design that I've got now, I'm not sure what I'm going to do with 76 tiles. If I have them all available for BV's brain. And you know, things will come. It's an important thing to do next. But that doesn't seem to be as important right now. So I'm going to push that off to a stretch goal in the last. I'm hoping that there will be, well, I've got all kinds of hopes. We'll see. So that's the schedule update. All right, the sensory motor stack. This is essentially like the spine and the nerves and everything that connects the brain, which is just this pile of goo connect, you know, buried inside a tough skull, deliberately isolated from everything outside as best it can, except by the specific stuff that's fed in. That's what we're doing here. And here's what we've got. And I'll just go through it really quickly. This, the bottom level here is written in C++. It's the bullet physics library. That's the simulation that provides, you know, check boards and squares and boxes and cubes and friction and collisions, all of the stuff that we saw in BV's adventure. The guy that wrote this got an Academy Award for it, you know, one of those technical Academy awards that they push off stage, but still pie bullet is a layer on top of that to connect it to Python. That's where I, so these two things were out there in the world. These four things, pie bullet time, Mr. State, packet spine and packet IO are Python packages that I wrote. I mean, some of them are pretty small. Some of them are a little big. I mean, essentially this is all just glue code to get us to the point where we can send the signals, the sensory data through these serial loop tabs that we made that I've talked about in the last couple of updates to get from the workstation. All of this stuff right here is currently running on the workstation that we're sitting at right here now. Once we deploy this onto the matrix, it'll be running in the dark room on a different machine, but you know, none of this is time critical. All of this has plenty of time. I was not worried about trying to pick the fastest way to insert into a Python list or whatever, but I was worried about how to get information from the host, whatever it is, to the T2 tile. And then up there, we got some more Python code. So the packets that get sent by packet IO get caught by packet IO at the other end and go into the T2 SerLoop processing stuff, which is also Python and then now goes to a new level, the brainstem level, which is written in C++, which is at the moment a stand-in for all the rest of the thing. So where we are so far is the sense data starts in the simulated world down in bullet. It goes up through PyBullet, all of this stuff. It goes over the serial lines. It gets up to the brainstem and the brainstem is where the decisions are made, where sensory data is converted into motor commands, which then work their way all the way back down and they get applied to the world in the next little time step, a 60th of a second, something like that, I think. So that's what's currently running. All of the stuff that we saw was doing this. The actual act of changing sense into, break into action was happening on a T2 tile. The goal is to not have it in the brainstem, but to get all the way up into ULAM code so that we can actually write spatially distributed, robust, all of that stuff to perform the processing that we just hacked in the brainstem. And these were the second generation just to close off that story. I screwed up the first one as I'm, well, software guy lost in hardware land. These second generation Serialoop tabs work fine. I'm totally happy with them. They're great. Okay, that's that. Now, the community, you know, I can't say enough how much it helps for me to have someone to talk to. So the live stream audience especially, you know, thank you so much just for that much. I mean, for carving out time, you know, in a specific space time coordinate. I mean, it's a big deal. Thank you. Also, we have a new nerd, AJB Kim, Living Computation Foundation nerd 253. Thank you for signing up. At the same time, I want to especially call out, we have a bunch of recurring donors, including something pretty significant, you know, and I don't thank them enough. And so I want to thank them today. And all of the LCF nerds for playing along for saying, you know, yeah, let's let's have my get my number in this story arc, whatever it is, however far it's going to go. The if one can stand to deal with discord, there is a discord that we have chit chat and so forth. It's pretty quiet. It doesn't, it's not that interruptive. But anyway, all of this stuff is wonderful. I mean, we may actually spend a little money soon. And I'll talk about that in a second. But one of the things about these donations that come in is, you know, I've been spending mostly my own money to do all of this stuff. And I want to put the money into the foundation. But basically, I have to be careful not to put too much money of my own money into the foundation, or the tax consequences are bad. So the more donations that I get from not me, the more that I can put in and still be okay with the IRS folks. So all of this, thank you. Yeah. And look at this. I mentioned this before. And I talked about it on the discord to the folks there. There was this, this new hardware called Grayskull from a company called 10s Torrent that has this new dev kit that, you know, I came across this thing here, Grayskull E 75 $599 for this PCIe card that if you go look at it, you know, it's got tiles, it's got intertile connectors all put together inside this thing. And it's got a whole bunch of hardware built in to connect these tiles to connect these cards, the Grayskull cards to other Grayskull cards through the PCIe bus and so on and so forth. So I sent the letter saying, Hey, I was interested. And they I didn't hear anything. I expected that was going to be the end of it. But they, I got them that email back and they actually, you know, Google for movable feast machine and saw that what I was proposing to do, what we are doing in the T2 tile project is related to the stuff they're doing. And maybe we want to go ahead and buy one of these to evaluate it. Now, you know, I've looked at enough things and I've bought enough things and tried them to be, you know, very cautious that they're, this is aimed at doing, you know, AI machine learning stuff. And so it's really certainly wasn't designed with a movable feast machine in mind, but it's a lot of computing power and it looks like it might be able to run different code so we could have element number one having an event on this part of the chip element number two of a different type having an event simultaneously on a different part of the chip and so on. So that could be pretty exciting for further out. So yeah, so there's the Grayskull and even look at this, you know, so they have two programming models, TT Buddha top down approach and TT Metallium bottom up approach. And that's what I wanted to talk about in the car run out of time. The whole building the here, I started from the bottom and worked up in this sense. And the whole point was to not let these things be coupled together. Oh yeah, I wanted to do a demo. Let's see if there's, see if there's time. So up here, the thing saying updates, that's running on the key master. And I'm just going to do this and then I'll explain what it is in a second. I'm going to change this value. And then I'm going to say make and now so that's the brainstem that's running up at the top. And there wasn't even a simulation there was no world. But so I start up the world and I change the configuration there it is. So those probably hard to see that means that the world downloaded the new configuration file to the brainstem which picked it up and went with it. So all of this stuff. So when we're going to be doing new simulations with the world, the brain is just going to keep on running. It's going to be like us playing a video game. When we die, we our brain doesn't get reset. Our character gets reset goes back to the spawning area, but we continue to think and that's what's going to be the case for the T2 matrix too. So it's kind of cool. I think it's kind of cool. So okay, and that is it. And so the goals for next time, March 5th, BV's brain coded in Ulam and running on the matrix and then BV changes his own wiring and then fun is fundamental. Have a lot of fun. Folks, thanks so much for dropping in whenever you come to see it. And I hope to see you next time.