 I did see you roll your eyes when I said that last yesterday almost a face bomb too It's all it's all of the video. I've been uploading the videos. I've missed a couple again because sorry community You know things happen Also, Tuesdays, I definitely don't come to the meeting. So there's never gonna be a DevSync video for Tuesdays But um, anyway, we've got a few out there. Hopefully we can keep the ball rolling Yeah, do you want to go fast? Sure This will be pretty short Lots of meetings today on our CM selection process But a good Kind of review of the the proposal from a company that has both manufacturing options in the US and and In overseas in Vietnam. So that was kind of interesting a good Process that they have to starting here and then moving over there And then we went over Some of those so the stuff that we've gotten so far. We're kind of really only about halfway through we've got more to go through next week We've got presented that to Michael where we're at And then other than that, so several several meetings there and then other than that we uh, I was just working on assets for the video for the The fundraising effort and that was about that was about it for me But I'm keeping the printer running for prototypes All right, thanks Derek Let's go to Chris there since he's unmuted already So another day in intent purgatory I the good news is I did find out why Asking for a timer was giving us a future time And I did also find out that the intent one of the intents in the timer skill was returning Pedacious if you don't use pedacious, right, you can get some really weird results. Let's put it that way So I've been talking to guys a little bit about you know, maybe And I think part of the problem too is that we've I mentioned this again just for the meeting we Promote pedacious as like the next coming of intent parsers and it's just not there yet You know either that or we just don't know how to use it well enough yet one of the two or both so I'm gonna have to do a little bit of intent Work not only in the timer, but maybe convert that one and predacious intent in the daytime scale To an adapt intent so it doesn't gobble up my timer utterances Because not only did the daytime gobble up my utterance it also gobbled up with a point nine six Confidence rating so it really really wanted it back So, yeah, so I'm I've learned more than I did last time I went through the intent parsers about intent parsing I still have a little bit to learn but I think you know I don't think there's anything wrong with knowing one of us knowing more and more about that stuff Yeah, I I've always been suspicious of mixing the two different ways of doing the intent parsing because I Just can't imagine a way that you could compare the scores between pedacious and adapt I just it does there's no there's there's no sense that I can make a bet because you know They're just completely different algorithms. Yeah, and one of the algorithms is explicitly Indeterminate like that that machine learning language, you know approach It's just it's never gonna be consistent with anything but itself Yeah, as I told guess our neural network as dumb as a doornail But I mean and So we've talked about things like using adapt for for keyword searching and maybe just you know trying to use pedacious for one-off type Type commands that might you that might mean the same thing as You know like instead of saying set a timer of 10 minutes You could say ping me in five minutes, you know That wouldn't fit in an app in the same in depth intent But it could be you know there for a patient's intent. So Anyway, just I wanted to bring it up here because people are watching some of this Well food for thought as you use your intent parsers to be careful because you know if it gets if you're right It's so great It's up other skills intense then we're gonna have problems On a big a big part of it comes down to that entity matching, right? So like yeah Yeah, you know this offset entity which is intended to catch you like 20 minutes So they can say what's what's gonna be the time in like 20 hours or something like that? Instead of just matching 20 it matches set the timer for 20 Which like is clearly wrong so If it didn't do that then it wouldn't you know if it was if it was properly Matching the entity to be you know only an integer then It wouldn't be able to match the intent at all But as it stands it matches it an extremely high confidence rating So I do think yeah, we need to have a look at that entity matching and per day just Sometime Yeah So hopefully I will climb out of this rabbit hole today and then on Monday. I'll start now, you know back on Testing can be new testament changes to the timer scale, okay? Ken So my neural network knowledge is dated but I Thought they required feedback to operate properly So if we don't it depends on what you mean you can train a neural, you know a simple You know classifier or something like that with just inputs and expected results like true or false Or something like that. Yeah, but the whole concept of it I thought was that it would learn from feedback Whether its guesses were correct or not and it would then Internally adjust the values that it returned based upon the number of positive versus negative hits over time for particular Entities, but who's gonna give it that feedback? It would only you could could only be the user and we don't have any mechanism for that right now Which leads me to question why we incorporated it unless somebody was intending to do some sort of feedback or Some sort of personalized Training well it goes back to the machine learning loop, you know And once we close it through a wakeboard the intent is to close it for pedacious as well You know, we just hadn't you know this wouldn't just one of those things that needs to be done Hasn't done yet. There's yeah So since that's not happening it shouldn't surprise anybody that pedacious is Performance is is not great So that's not really surprising but I've been working on the builds because Chris and I found a timing issue in the bring up sequence where we can get out of sync With the panic or back end and never recover And it emanates from the fact that we basically give it one try and If it fails it fails and we don't know if it fails or not so That's should be addressed at some point I've been able to replicate it and it's definitely timing related because The only way I stumbled upon it was to burn the image onto a USB 2.0 drive instead of a 3.0 drive Which is significantly slower and causes different behavior regarding timing and It consistently Produces this error. So that's your steps to reproduce Simply take yesterday's image or the updated image from today, which I've done and both of them behave the same way burn it onto a 2.0 Boot off of it go through the process. It'll eventually pair with the back end of Selene It will also eventually sync with the back end of Tantacor, but Selene and Pantacor don't sync up Well, that is an awesome a way to reproduce a potentially very tricky bug. So congratulations there. That's that's great Yeah, and then I also took the time when I was out at the store today to buy a 3.0 drive So I could get the latest build working. So that's what I have An extra kudos for the practical solution so you can keep working Hahaha What if I'll eventually coming back with another 2.0 drive? Oh, this one doesn't work either gee Just be a lesson boys and girls if you have a mark to don't be by a USB 2.0 to go with it Well, yeah, that's a solution. Um, so anyway I'm back on converting wiki to common play and then Gez was kind enough to share a PR from Jarvis that updated common play more along the lines of GUI support Although rumor has it there might be some other substantive changes in there So once I get the wiki skill Ported to common play then I'll look at that PR and do my best to get that incorporated Yeah, so this one this one's just like ads video playing capability. Um, so it's not it's not broader common play kind of stuff. It's a nice little contained one Um Yeah, and to be clear about that panic or issue basically we if I don't get If the device doesn't get the device ID from the panticore device ID from panticore before the whole pairing Francis is complete then it Reports to Selene like a non panic or device because it it doesn't have that that's what I depend on To know if it's a panic or device or not that has a panic or device ID So because of the slowness if you didn't have the panic or device ID until after pairing was complete Right, right, but yeah, the thing is that I'm pretty sure all mark to should have a panic or device ID. Is that correct? Not necessarily Panticore-based mark twos. Yes. Yes Well What mark twos do we support that aren't panticore base? Not that we support but like you can run any stuff where you like on the on the device Sure, I got that So that's me and Michael you're gonna be out starting Monday or Tuesday, right? I was gonna go on vacation Monday or Tuesday. I'll wait a week If you'd like I can wait until yeah, why don't we talk about that off offline? Okay, so that's my that's my status. Okay cool Yeah, so yesterday I did a stable release of the of the mark to so that everyone got 2102 because What I forgot to think about was that our lovely update process started bugging people to to manually update their devices Which you know should have thought about but anyway we Need to fix that before the next major release And I've been Trying to get through some of the PRs on Minecraft core some of them were explicitly, you know set to to wait until the major release others have just been You know in the backlog for too long So I've done one for the AVX check and Just started like flagging a few that particularly if I've needed to work on them then you know that the policy has always been that we Have someone else approve them or not so Yeah, there's a few with with Chris's name next to them, and there's a few there's a few with kens, so That will help us help us to clear the backlog a little bit and Yeah, I think I think two major did some more recording with Stratus The get response video took much longer than expected Because it is a little bit of a Like it's such a simple method, but then you can get pretty detailed in terms of validating responses and that whole process so Yeah, that that should be coming out at some point in the future with we're trying to bank up a you know a Number of videos so they can they can consistently come out every week Which they come out weekly on Fridays it in the afternoon in the American afternoon, whatever whenever that is Yeah, anyway, that's that's about it I think okay Well, I have not been working on dev stuff, but a little update from the business side All of our fundraising stuff our paperwork is more or less complete. I'll be checking in on that in just a few minutes But we have commitments for significant sums of money at this point which is very good and And we'll be launching phase two of that is currently scheduled for June 17 And so so yeah a lot of good news there. I'm being intentionally cryptic about that because we're not allowed to talk about it publicly just yet but But yeah a lot a lot of good news there Significant sums of money is always welcome