KZG Ceremony Breakout #4

Transcript

hey hey welcome hey everyone howdy daniel can i get somebody to talk to so i can

test my audio test tests thank you thank you [Music] great let me um let's wait one more minute and then get started but this is probably all the people that will be showing up me get the agenda i'm doing that thing where people narrate when they're leading a call so don't it's nothing else fantastic there with you

also i apologize involved for uh bad audio quality on the cafe in paris how is paris is it so hot as everyone complains about yeah we melted we melted it but uh two days ago but today's today's not too bad conference five's definitely shifting i feel like in a post uh dev connect worlds there's like now technical conferences and not technical conferences and this feels like it's it's strap straddling the line but it's by thinking what's not technical at least in comparison to

previous years so that's a bit interesting you mean it's hype yeah there's just more more hype and like the real work gets done more and more inside events like the the the people i want to talk to aren't even bothering to attend the conference or don't even have tickets yeah i had the tickets were like extremely expensive this year yeah and super limited but frustrating all right um i think we can get started there are some new people on the call do

you want to introduce yourself okay so to start this is the kcg ceremony call we're building uh we're crafting the trusted setup implementation and the coordinator software that's going to make kcg's trusted setup possible which will provide a foundation for thanks sharding prototype sharding sometime next year um so we're getting started early on this in advance of when we'll need it um yeah it i see there are two new names you want to just give a quick intro to who you are and what your interest is for the project okay so hi i'm daniel uh some of you

already know me from eve break where i implemented together with mario's um our small prototype of the specifications provided by carl and i'm here to catch back up with everything that's been happening in the meantime yeah great welcome glad to have you alexa are you able to yeah hello guys i am from netherland uh pro developer uh i'm here due to some interesting points in agenda i like to follow our other clients updates so maybe

it is what i want i don't know it's like a first time like yeah it's great so you're just here to learn and see what it's all about you're more than welcome to listen in all right why don't we get started with just general updates from the last two weeks what people worked on uh general state of things any major updates carl let me start with you sure on my side has mostly been more worrying about uh things like how do the satchel stations work uh what does that look like uh i don't know if i've discussed on

these polls before the idea of sort of linking attestation back like double linking it and then also worrying a little bit about simplifying the api for participants who don't want to implement the full api i've sent some messages about that um that's that's mostly what i've been thinking about slash working working on but uh i guess those are less updating things and more researchy things we can discuss later sure yeah and i realized we forgot to give guillaume space to intro himself so guillaume why don't you just jump in and intro um yeah hi i'm guillaume uh core dev at

the indigo ethereum team i'm interested in uh potentially writing a zig implementation because i figured wasm would be a nice target so i'm mostly here to figure out if daniel is going to proceed with his rust implementation uh yeah so that's roughly uh my goal and interests awesome is this going to be just for you or are you going to you know do the implementation with a group and run a bespoke ceremony with the group or just you uh yeah assuming uh there's no wasm implementation i would do that uh i mean with a group it's just myself

yeah but i would i would be interested in joining if if i ride the software cool uh jeff did you have any update on things you've been working on okay uh yeah sure um yes uh so i've progressed the api specs to some extent got some feedback from carl we [Music] did release ask for more

contributions on our test site and that raised a few issues that we have to solve like it doesn't work in firefox and it doesn't seem to work very well in mac um uh what else i have a developer started a couple of days ago who's hoping he's going to be helping out and for the time being he's just catching up on the tools that we're using and that kind of thing he's got a lot to learn so um yeah that's what's been going on is he uh or she this developer are they

uh part of the pse team or did you hire them yes the contractor for this okay cool he's well he said he's doing a uh yeah like a three-month trial period i guess but uh yeah he's basically on on the on the pse team um see anything all goes well yeah great that's good to hear um you mentioned it briefly but do you want to talk more about the test test site i did try to do it myself chrome on windows but i think it never actually went through um is this something that

you want to get more people to test or is the small group enough um yeah we've only had a few a few trials successful trials so yeah i think this the system is a bit clunky like if someone's contribution fails it doesn't i haven't got all the bits and pieces in to recover from that um so a normal contribution would would just be ready to accept the next one but i haven't put the failsafe in there um so yeah it's going to be a bit clunky i've got from those who have contributed we've got maybe

uh you know six or eight or ten valid contra contributions which gives us an idea of the range of of you know of of what the performance is going to be and and importantly we've hit a few of those those configurations that don't work that we know we need to work on trent you mentioned that you are unable to participate in this so that your participation failed do you have more info on what happened there nothing beyond it just kind of

i think it's i don't remember the terminology it just said it was basically it's it seemed like it was always stuck in the startup phase or maybe that was doing the computation and it for some reason never uh concluded or at least as long as i kept checking it it never concluded i can try i can try it again so there's i i encountered a simple thing like this and had to have to refresh um on one one of my my runs so i got through the queue got to the front of the queue and then there was sort of a like preparing or something phase before i could doc before it even downloaded the transcript and it just never got through that okay

and then i refreshed and it instantly downloaded because i was still front of queue et cetera et cetera so um yeah yeah this is generally not going to be a queue and refreshing would always sh yeah it should always just pick up from and try again okay um i the the the logs weren't particularly helpful and the networking reflection was something that everything wasn't helpful either in terms of seeing what went wrong so i don't actually know what the failure case was but i think i could just look back to trent okay um jeff can you okay so does this represent

sort of the latest state of things the the the test site um how far off is this going to be from the final implementation oh um so finally implementation um there'll be a lot of different stuff behind the scenes there'll be the api that's got to be built um the the actual user experience probably not all that different yeah yeah all the most of the works behind the scenes and and clearing up those you know if if we're getting error cases then we have

to clear all those up so that we get um [Music] but yeah we get all those edge cases cleaned up um do you have a link to the code for that website code for the website yeah i think those links are in the telegram channel um both the with the test site itself and the code let me uh draw oh uh drop it in there anyway so

so one of the other things i saw when i was uh playing around your thing here jeff is that it talked about uh a p-tower file which is the standard used by some other ceremonies um is that just what console console log is saying or are they actually stuff being converted in and out of the ptal format behind the scenes um it's it's just using that uh well no it's just it's not even it's just using a file that happens to ends in p tower because that's what the previous implementation was using

um and it's not using your your um format it's just it's the for the data is the original version of kev's code so it's just a bunch of bytes there's no json structure ah okay okay i see and then you're just calling that that that by construction a p-tile despite not being the same thing as other p tiles that exist and also not the json yeah okay yeah that's right and so then there aren't any major the reason i mentioned the anti-major

sort of um computation time or speed differences between that and the json modulo the json encoding it's not like we're re-encoding into a weird format uh won't have that much that won't have too much impact on the performance at all okay cool so it seems like the goal will be to just keep iterating on this finding the edge cases in various combinations of whatever os they're running and hardware and just keep ironing those out yeah nice

is it too early to talk about how long it's taking on in the average case oh uh no um we have seen five second and ten second contributions um they uh i'm guessing that's a typical one is going to be about 10 seconds i mean that's that's great that's much better than my experience of not starting so i was worried yeah

one interesting point on that though is i mean carl asked for um what what's the typical range of of times we see from previous ceremonies so i did some analysis on that the comments are in that document where carl was asking the questions but um there is a a very large range it's a very long tail like a lot of people have have a reasonably short contribution time but you can go up to ten times that time and you're still getting the last five percent or ten percent of the of the distribution interesting so it's both long tail and

fat tail the distribution yeah huh okay this is some implications in terms of of how we set time marks yeah if i ever just say 15 seconds and we need whatever 150 seconds and even that might not be enough to capture capture everyone um based on i mean i assume you don't have numbers for this so just based on your intuition do you think this is due to people using uh their own slash different implementations or is this a hardware thing yeah this is all in a in a single implementation so it's purely the hardware or whatever platform factors come into it

um i did a separate in our analysis for internet times upload download so that's that's excluded and it's pretty much just the hardware variations yikes okay that's interesting and is this running um wasm in the browser it's what it's is this what's in the browser okay yes because i i feel like a 10x a 10x spectrum over hardware is kind of surprising but maybe not over wasn't implementations between different browsers or something

yeah [Music] this is multi-threading makes a big difference if um uh some hardware will allow four threads some will eight more um uh so [Music] yeah well i mean i did what i'll do is try and link the the statistics i'll drop a link to it what's what's the verification time on the server like yeah we we haven't got that set up at

the moment i'm expecting it to be just three seconds or something is that fair yeah i would almost expect less than that it should be it should be blazingly fast on the server side um the the big difference between what you your implementation marius and this is that you can uh you can batch the pairing so it's only one pairing there's one one pairing in total plus one multiplication per point sorry one one multiplication one addition per point um so it'll be very cheap [Music] yeah it makes sense and the

verification is done in rust with kev's implementation right uh that's that's the plan yeah okay morris do you need uh are you guys going to try implement the the verification as well as then would it be helpful for like do you know how to do the the batch verification there would be helpful to have a doc explaining that for you guys

i'll put something together for you guys thank you anything else on the test trusted setup test site okay somebody guess we've covered it all oh well i see the sheen is here

do you want to just say hi hey guys how's it going i'm just dropping in to kind of keep an eye on the progress of this uh you uh machine works with oboe if people are familiar do you want to just give a short intro yeah absolutely so oboe is one of the teams working on distributed validator technology and specifically we're making a middleware client that sits on top of validator clients and makes them highly available so you can run a number of kind of client combinations together and have a fault-tolerant validator client and we're just about to kind of launch test nets and stuff if people are kind of i don't know keen to try out new

types of validators cool glad to have you cheers thanks john um the two things going back through the chat well there's the api i think we already talked about jeff had written some stuff up about that but there carl you had mentioned doing an alternate queuing system do you want to talk about that um just summarize it and where you landed on that yeah sure so um as usual i'm just trying to maximally optimize for someone to write an implementation themselves and to do this as simple as possible and although i really like what the

queueing system's gone i think we can make it even simpler and so my idea to do this is to have an alternative form of queuing that exists at some time where we switch from like this this interactive check-in queue to a system where you just you you call the api and it gives you a slot of a longer period of time and this is for people who want to use their own implementations or people who know they have really slow hardware or trying to do some weird stuff with sgx et cetera et cetera it's sort of distinct from what i'm seeing as the um like intro the interesting experimental um

uh contributions like that'll be later and that's more about the ceremony around the randomness and destroying it this is just for different implementations um in the message i sent i was proposing saying like maybe every eighth day we do this but uh i've like since decided that's that's a little bit ridiculous and makes the ux very hard to to come around on so maybe at the towards the end of the general contribution like general public contribution or at the beginning then we should we should have a period like this so what that means is it basically looks like a different api for a different api endpoint for the queue where you can just call it and it'll say okay here's the next available slot and then you get that

does anyone have any thoughts queries comments on us sorry can you quickly explain it a bit more so you you get like basically tickets what's the what's the difference either we lost him or he doesn't realize he's still muted sorry then realize i still muted um the existing system you have to sort of you check in and then it gives you a new

time and you have to check in by that new time and then once you've done that it gives you a new time etc etc until we get into 30 second intervals right before your your contribution um the idea for this is that if people are taking really long or really short to contribute in the queue in front of you we can shuffle the queue around a bit to accommodate this and hopefully achieve higher throughput um this is really good for the like general contributions we're a little bit worried about speed uh so a little bit worried about how easy it is to to implement so i'm wanting to offer a separate endpoint where you will apply and it will give you a time so it will say like come come back at 3 pm tomorrow and then there's no more checking in you have to do you just log in at 3 pm

tomorrow and say hey give me the uh the latest ceremony it gives it to you and you have a fixed time length to to participate so we sort of switch from this like dynamic queue to a system which is like reserved time slots um yeah okay and as closer you become to the um specified point in time um the intervals in which the participants are supposed to poll the current state is getting shorter that's for the that's for the normal queueing system

but for my alternative q system there's like no for the chickens has there been some investigation into something like um having the participants connect to like a websocket endpoint so they do not have to pull but instead the the coordinator pushes the i say let's say um current current uh transcript to them as soon as they are supposed to uh yeah yeah so this this was a bit of a bit of a discussion i think uh you've

missed in the past and that's like we're trying to move away from this too to coach and quite simplify the api um the like the checking version is not extremely complicated or it's not aggressive pulling like it's going to be if i start every few hours and they get down to every 30 seconds um so it's nothing the server shouldn't be able to handle all right um yeah it just it makes it really easy just to to to be able to use standard um like http get requests right a little bit easier than a websocket [Music]

anything else not for me at this point unless people have any ideas as to whether this is a thing that's necessary uh i'm sort of torn on it because it's not actually that complicated to just do the whole interactive system to me where it it shines for maybe people who want to have an air gap solution or something like that um yeah that might not be necessary yeah uh i i like the idea i think it's um it'll it'll be easy to um to switch and once we're over the initial

mad rush um there won't be we won't expect long queues or um so yeah we get more flexibility in [Music] and we can cater for all those experimental more experimental implementations sounds good to me yeah um it's also like if we're running out of time on the implementation front we can implement this after the system's already gone live like we can add this to the other queueing system once we've got the primary one out of the door one question

i have is how the authentication is currently done with with the other system because if you have this periodic check-ins and you have to re-authenticate for every every check-in and that's uh yes it's it's two and they'll have a token the token will be validated on every request but it it'll last for a good length of time so i don't think you'll need to be renewed this is just a jwt token right

yeah okay so currently on the side you can connect using your ethereum account is there some criteria that have to be fulfilled by that account so um or or can i just generate more accounts if i like and use them to participate multiple times yeah so um we are talking about criteria uh like such as um nonce greater than or equal to three some people have floated other ideas like having a pull up or at least

one leaf in the account so some reasonable barrier to entry for sure all right thanks i would like to start having a place where we can discuss these these things a bit further exactly like what are these contributions like what are our auth methods uh just for like general general discussion um i've had a few people like come to me quite worried about what what are and aren't the criteria so if we could have this in a like easy to access public case that'd be good um so jeff you wrote up some of those those your ideas originally in that

laptop that's that's that's pretty good but i'd like to like put this into a repo or some other public forum that's a little bit easier for multiple people to contribute to and have discussions about um and also just a place where we can point others towards uh if they're interested or have ideas on what should be included sounds like kcg spec ripple to me i agree let's add it um okay subfolder for for api maybe um yeah yeah are these i'm assuming these are

people that aren't already attending these calls so they're more than welcome to join here as well and have that discussion it seems relevant and we would make space for it yes no they're they're welcome on the calls it's it's more just that they're like not really wanting to to fully follow this process but they just have some opinions that watch initially included and i'd like to accommodate them uh also just have a canonical list and to actually have the debates like does the non sequel three five or two kind of thing and like just just have this out loud have this in a easy-to-follow place yeah no i think that's great uh the more we can do to open up discussion to people the earlier the better

um part of the reason we have a uh public discord channel is because we want more people engaged and asking questions and figuring out how they can participate because that's gonna yeah it'll help us in the long run uh and this is something different to what kev had mentioned last week he was talking about another uh let me see what he said another one uh he was talking about making another repo with rationale type documents oh yes yes this is this is separate from it okay one is sort of understanding the general ethos of the the ceremony and then this what we were discussing moments ago is

like the specifics of how and why you you participate philosophy versus implementation i guess yeah um does anybody know if he worked on that or has an update not for miami okay then i'll just reach out and see uh i'll dm him uh yeah and then you also mentioned the the sdk is there any update on that uh i've sent some stuff hold on back back with kev i'm just trying to figure out exactly what that looks like but

nothing nothing nothing major or helpful there um i mean what what this actually looks like is basically just like what once cam's happy like kev and i are happy there's i guess not too much discussion on this unless jeff has any feedback so um like it doesn't affect the wider implementations for now got it um let me see okay it looks like i'm just going through the things that we said uh two weeks ago we were going to take care of so you had said final decisions on api

um sounds like that had some progress i don't know if we're at like final final decisions but um you made progress on it right yeah uh jeff skipped the stuff jeff did was that was pretty much that i'd like to now sort of just integrate this into the repo um and then we can yeah lots of discussion points there but like essentially i i'm pretty happy with what's good awesome is making the spec implementable the sdk or is this something else nope that's something separate and

something i did not do okay so that'll be on your something you want to try for the next two weeks uh no i i unfortunately i i'm away for the next two weeks um oh right yeah not not all of it but enough of it that i can't i'm not gonna make promises about this okay um okay uh i think that's most everything i had from last week and general comments that i saw in the chat

or general discussion points that we can pull forward there's that uh four eight four four i think that's the right number the prototank starting call on the 29th but i'm guessing you're going to be away for that carl so i'll try to join and just provide a small update um but yeah we've gone through everything i had on the list it sounds like there was definitely some good progress this test site was really nice to see that we're getting it out putting cobbling together the pieces um and figuring out how it's breaking what's good to do earlier um carl if you

is there any update to the timeline as you see it are we still on track uh anything that you're worried about as we're heading towards the audit in september or august i always get this messed up um uh i guess okay so the second half of the second half of uh august is what we have yeah so i guess that's that's um the the limiting factor here is going to be the sdk um and that's i guess like that needs to be fully like defined and implemented before before then uh like

in in terms of the actual amount of work that needs to be done there's not very much but like there's a bunch of agreement stuff on like small things that need to be sorted um so i'm a little bit worried about the about that right because it's basically a little more than two weeks two weeks away um is there anything that so i i it sounds like you're trying to yeah you've got a lot of things on your plate at the moment um is there any way we can get is there anyone we can get to help you with sdk or is this something that you just need to do yourself well you're going away for the next two weeks so yeah that that that's that's my

issue right um hmm okay i'm just i'm i'm i must think about this um we can check should we control the sync things um yeah all right we must do something about this yes um is there anything else on the agenda we didn't cover or sorry something that's not on the agenda that somebody would like to bring up or something i missed from the agenda um i opened up a discussion with our psc designers

because i figured um the design thinking could happen in parallel with development and spec development particularly thinking of things like a dashboard so we can display progress and um uh [Music] so the designers our designers got in touch spontaneous with the ethereum designers and they seem to have formed a you know team to work on it but how how do we want them to how do we want to get engaged with them going forward do we want to

should we have them on these calls or that's the best way to do it um this seems something maybe more suited for my non-uh cryptography skill set so i'm happy to help here engage with them and provide feedback on the interface stuff like that um yeah definitely at a minimum i'd love to have them engaging in the kcg ceremony channel and then yeah they can give updates here as well i don't know what time zone they're in but yeah they're welcome to join here okay um i can i'll uh

message you after to get there i i'm assuming they have a chat somewhere where they're talking and collaborating so we'll just figure out what that is they have no i have added them to the discord so that they're present in there is the two pse yeah i think it's let's combine for now and uh if it gets too much or the discussion like we run another time then we can split up later but uh yeah pronounce that that seems to make sense all right um

i think we're kind of tapped out on things to talk about unless there's anything somebody wants to bring up yeah so i'll be uh i'm sorry go ahead oh it's just gonna say i have a speaking slot at east seoul uh and that's early august i was gonna talk about this and flirt a few um hackathon ideas um cool yeah yeah you know yeah that sounds awesome please go ahead um like

part of my push to keep this so simple is that this really should be doable in a hackathon so yeah that sounds fantastic yeah the more implementations the better um so the other point i want to talk about as i i've been having this discussion with several people on the idea of doing uh sort of a pointer back from your uh from the transcript to your identity right so the the the standard way you attest is you go on twitter or your favorite at the station platform and you say this contribution number 27 was me um and that's fine but what happens if someone else claims to also be

contribution 27 so my proposal for this is that we could have what we can include in the the transcript itself a reference back to some kind of identity for you be that your ens or an address or something like that so the transcript itself would be contain a signature of say called beak.if and then i would also claim on twitter or with the transaction that i was contribution 27 and now we have the sort of double back to this double link system between the two uh which is which is quite nice um in terms of the actual security model it does not change a thing which is kind of interesting so if you are prepared to trust uh you trust marius and marius says on

twitter he has contribution number 69 then that should be all that you ever need for for the trust model it doesn't matter what anyone else did even if even if someone else claims to have the same contribution if they point to the same one as long as you trust marius then it's okay um so i had mixed feelings on this and then i'd also add some complexity and adds a little bit of space to the transcript but i did the math on it that's that's not too complicated um but i've subsequently decided against this idea um primarily because we've decided to do the uh the node also the um the subgroup checks after the

contribution um and this would have to be done before you can do the subgroup checks which would mean that the contribution would contain a signature of uh say carl beat.if before you've done the no degree checks which makes it harder to like deny that there was an acceptable contribution so i'd like this idea less and less now i don't know if i explained everything well enough was this your optimistic evaluation or is it something else yeah so the optimistic evaluation is the sort of the reason why i don't think this could work um that's an idea that like hasn't been done in any ceremony before and as i

said like doesn't add anything from a actual security standpoint but this double linked idea does add a bit more like authenticity to contributions that we would see it would like help people trust that i think if you if you could see these double links even if like statistically attribute but it would also make attributions reputable yes i mean it would be purely optional right you could you could still contribute without okay yeah i don't yeah i don't think it i think it adds complexity without adding anything worthwhile

yes um so the the complexity itself is is basically free uh because you already like it's it would be a bls signature and this is all over the bls the bls curves anyway and the signature is just one multiplication and so the only quote unquote addition to that you would need is the hashtag which for like if you're using someone else's vls library will already come included but if you're running your own library which actually should be pretty doable then this is quite a bit of overhead so that's i guess another another point against it okay so i'm in a thin like yeah

i was just gonna say is there a write-up or something that we can see all the arguments for and against this just i mean i know i uh i i think issue number one on the specs repo refers to this one of the low numbers like one or two um refers to this and uh like it does it it does before the optimistic idea so um it doesn't include that argument against it but uh it should be there um but i think more the reason i'm raising this is i want to say that i've sort of gone back on this being a reasonable idea in the

first place and so just for the public record i think we should we should stop doing this this idea got it okay yeah uh that's it from from my side in terms of the new points of interest anybody else any other topics okay i think we can probably end early this week

um just we can end with what people plan to work on over the next two weeks um obviously the audit is coming up and carl is going to be away for part of that so we'll figure that out jeff what do you want to get done in the next two weeks um try and make some progress on uh the the the issues with the competition there's failing cases uh i i think my next two weeks are going to be fairly uh busy i'll be traveling in about a week

um so i'm not sure there'll be too much other progress but nico is is working on on issues and i've been converting you know assembling a list of github issues for him to work on so there'll be some progress there great um is there anybody that you i mean vitalik was in the chat talking about some of the participant id strategies is there anybody that you feel you'd like to have review these strategies before we move forward or like engage in the discussion

oh no so i'll i'll um i'll put this into a repo as carl suggested and um yeah then people can contribute in there okay so i think that that's the main thing is it's not necessarily the specific people that we need to contribute to we just need to have have the discussions in public uh make sure enough people have had the opportunity to voice their opinions on this um and then decide on exactly what is the list of methods here okay yeah i can that's definitely useful to have them uh in a repo that everyone can see and will that be all of your so i think you have like three notion pages are you

gonna move all three of those to uh yeah okay awesome all right uh anything else jeff or is that it no okay uh on my side we probably have some paperwork we need to do relative to the audit um i'm going to try to interface with the designers that are getting up speed and see if we can start sketching out some ideas and just figure out how they're going to integrate with the process um and

i guess those are yeah those are the two main things that i'll be doing over the next two weeks relative to this i don't know if we mentioned it but yeah we're also trying to there's a chance we might do um assuming everything goes well we might do a live contribution at devcon so i've been talking with the organizers about what that would look like how much time we would need things like that just making sure they're aware of what this project is and trying to see if we can get it get a slot on the main stage during kickoff i think that would be really great for awareness and getting people to engage in the project

maybe i already mentioned that i think it's pretty cool yeah i agree that's that's that's awesome yeah we're thinking of doing it is uh well it depends on what the the random inputs are and how that's figured in but um something like taking a picture of the [Music] the main hall and then that's the that's the randomness input yeah i think no matter what we end up doing it's not going to be actually a very meaningfully secure contribution but it's more just for the fun and

ceremonial but um that's cool too right right yeah exactly we're not pretending this is the be all and end all and yeah we can try all right um we can probably rap here i will connect with kev um just see what he was up to for the uh since the last meeting and then um yeah and one thing i mentioned to carl and maybe also on the last call is that uh ideally we start to have i mean last week was good but we want to make sure we're having discussions in public moving stuff to repo is also really good but the more discussions we can have in

that public discord channel the better for people who jump in a month or two from now and are trying to catch up uh yeah these are all good steps forward but we just need to make sure that any discussions that we're having are try to try to default to public just because it's it's easier for everybody to catch up when the time comes all right um honestly anybody wants to jump and interrupt me we can wrap here thank you everybody for this is the fourth call i think thanks for joining the fourth ceremony call uh this early planning stuff is crucial to you know making the ceremony work at all and in long term

the producting sharding and things other constructions like that so it's important work um it might feel like a like just another bi-weekly meeting but this is this is really crucial stuff so thank you all right thanks thank you take care thank you [Music]