Copy of June 2026 - Panel
===

Paige: [00:00:00] Welcome back to PodRocket, a web development podcast brought to you by LogRocket. LogRocket provides AI first session replay and analytics which surface the UX and technical issues impacting user experiences. Start understanding where your users are struggling by trying it for free at LogRocket.com.

Hey, everybody, I'm your host today, Paige Niedringhaus. I am a lead platform engineer at the startup AllSpice, and today we have a panel episode where we'll be talking about Cloudflare acquisition of Voidzero and rising AI tooling costs. But before we get into it, let's welcome our panel. We have Jack Herrington, a principal developer engineer at Netlify, host of the Blue Collar Coder on YouTube, and also the co-host of the Frontend Fire podcast.

Jack: With you, Paige. ha. Glad to be here

Paige: Glad to have you. We have Paul Mikulskis, a PodRocket host and a YouTuber. Welcome, Paul

Paul: Hey, thanks for having me. And ~ And it's been too long since I've made a YouTube video. Like, I'm talking probably two months at this point. Um, that's because we're busy working on projects, right? And that's what we're gonna talk about today. So excited to be here.~

Jack: ~Yeah.~

Paige: And of course, we have Noel Minchow, a developer at LogRocket. Glad to have you, Noel.

Noel: [00:01:00] Yeah, happy to be here, Paige

Paige: All right, so we have some interesting news. ~Uh, ~once again, Cloudflare is in the news. It continues its acquisition of basically all of the open source tooling and framework that is available today. ~Uh, ~so on June 4th, actually, Cloudflare completed its acquisition of Voidzero, which brings the entire team, including the Vue and Vite creator, Evan You, under the Cloudflare roof.

So Vite currently logs over 130 million weekly downloads from, and underpins major frameworks including Vue, Nuxt, SvelteKit, and Astro, which Cloudflare also acquired,~ uh,~ earlier this year, TanStack. Have not acquired TanStack yet, but who knows?

Jack: It is of the o- few remaining big open source, it is certainly, I would say, the only remaining open source framework project that has not been acquired or acqui-hired, were

Paige: So far.

but both parties have emphasized that the projects will remain open source and vendor neutral, and Cloudflare has committed a million dollars to an independent Vite ecosystem [00:02:00] fund. So ~the questions that we can kind of start off this part of this section are,~ Avenue has reportedly had a successful exit here on a product that was questioned for VC funding from the start.

We all ~kind of ~wondered, how are they actually going to make Vite profitable? Because it was open source, because anyone could use it, and because everyone loved it. So what do you think it says about the open source sustainability that acquisition is really the only viable path forward for maintainers?

~Is that, ~is that what it's come to?

Jack: I'm just tripped out on the idea that you just said everybody loves it and everybody uses it, but yet they have no path forward to profitability. ~And it's like, is that where-- I get, you know, that is it.~ That is like ~the, ~the problem with open source development is that there's just, yeah, there, there's only a couple different exit strategies.

You can make your own service, you can get acquired, or let's, yeah, I guess you can sell subscriptions or like a pro thing. There's just not that many ways to generate revenue

Noel: ~Yeah, I think it's, it's particularly challenging, like, it's, ~it's hard for me to envision a world where,~ like,~ there could be some kind of support service in the vein of,~ like, the old, um, you know, ~like the older open [00:03:00] source firms,~ um, that kind of, ~that kind of model of, like, we will, we'll have open source and then we'll have support for enterprise customers that are using this.

~Like, ~that just doesn't feel like that really plays nicely with these web tools that kind of exist exclusively in the developer tooling space. ~Like, I don't,~ I don't think ~that ~that model's really viable, so that takes that off the table a little bit. ~I mean, ~I don't know. I feel like in my ideal world there's-- We've gone through several proposals and nicer alternatives where there's,~ like,~ some kind of crowdfunding or,~ like,~ bubble up funding with tooling, but none of those seem to have really taken hold.

None of the,~ like,~ higher level abstractions there versus just the, you know, having tiers of sponsorships and the Buy Me a Coffees and the Patreons and all that jazz.

Paige: ~I mean,~

Jack: Yeah, coffee's only get you so far,

Noel: yeah.

Paige: it used to be that if you were doing something like this, you either had some sort of a hosting platform that you were going to offer that was the best way to use your open source tool, ~you know, ~like Vercel and Next.js [00:04:00] or, ~you know, ~different things like that, or you had a paid model that was the pro version.

~So, ~like ~ um,~ with Tailwind, for instance, they used to have,~ like,~ the Tailwind pro building blocks and things that you would pay for in order to get even nicer snapped together Lego pieces. But with... It seems like with the advent of AI especially, you don't necessarily have those, have a need for those things anymore.

~You know, if, ~if you're using a framework that doesn't play nice with your preferred hoster, you can have Claude c- code port it over in a week or, and make it play better, or you can, ~you know, ~have it read all of the things that the pro version of Tailwind has to offer and just build it itself so that you don't have to pay for them to give you those blocks to use.

~I mean, it's just... ~Yeah,~ it,~ it seems like a much harder road to go than it ever has been because AI can do it cheaper and faster [00:05:00] than we've ever been able to do it before

Jack: ~I mean, ~Yeah, I mean, Tailwind is a good example of that, right? Tailwind basically went from being so cash flow positive that they literally had no idea what to do with all the cash, which is an interesting problem to have, to literally the next year being essentially laying off folks ~and, ~and being essentially, ~you know, ~going towards bankruptcy.

And yeah, that's pretty quick, and it's-- and that, that was primarily AI, right? There was no need to ~get any of, you know, ~get any kind of pro subscription or buy their book, which is one of the- their big revenue sources. And so ~they just, you know, send-- they, uh,~ they ended up getting not even acquired, but, like, more just a large sponsorship from Vercel, ' cause Vercel realized that,~ like,~ "Yeah, we ~can't, ~can't do without,~ like,~ Tailwind."

Paul: Maybe it's also just a centralizing effect on what can be a viable open source, 'cause before it's ~kind of ~like anything could have gone. If do computer data thing with [00:06:00] a cool data model thing, there's a bunch of ways that you could charge for that,~ like,~ like you've talked about Paige ~in, ~in Ngix,~ uh,~ But there's open source out there that is doing very well, and that's still open source, like NADN, N-A-D-N, for

Jack: Oh, Naden. Yeah, yeah, yeah. I didn't know that was what Naden w- I didn't know ~that ~that was the

Paul: Yeah, NADN.

Jack: Naden.

Paul: ~ uh,~ Its, uh, valuation doubled since October 2025 two-point-something bill to like five-point-something billion. ~Um, ~and they're still open source, but ~they, ~they're what I would call like faux open source, where it's like you can drive the product to do some things, but the moment you wanna do like actual work with it, it's all in their like EE package.

Package is a /EE, their enterprise package, and that's how like their license they can say, "Ah, you wanna pull from that open source code? ~Well,~

yeah, let tell you, you gotta get a license for that." ~Um, ~so yeah. ~But, but, ~but it's working. They found a way to give like normal people who wanna do drag and drop workflows and maybe work with agents something they can self-host and like~ like ~do as many actions as they want.

~Um, ~[00:07:00] but the moment you wanna step into something that's ~like ~not self-hosted running your computer and a bit, bit more powerful, you need to pay the enterprise license, and it's not necessarily cheap, and they're doing ~very, ~very well. So maybe it's like a market fit thing for what could be a collective open source effort that gets funded at a actually like B billion scale.

Noel: Yeah, ~I mean, ~I feel like ~that was, ~that was ~kind of Paige's, ~Paige's point with the, like, having a hosting platform that your thing can evolve into. And it's ~like, ~okay, you can run it yourself, and then we'll give you the bells and whistles and the really nice place to play if you wanna jump in our, into our ecosystem.

I, I feel like the interesting comparison is,~ like,~ the,~ like,~ Red Hats and Agalias and all ~these, ~these tools that are ~like, this old, like,~ the old school open source mentality, which it feels in my head is more like, ~you know, ~how Vite and these other kinda, like,~ like,~ engineering build tooling steps,~ like,~ it f- that feels like a cleaner analogy.

But again, ~it just, ~it just seems like there's some fundamental difference in, like,~ like,~ the new web world versus,~ like,~ old enterprise, old,~ like,~ enterprise software that's ~like, you know, ~[00:08:00] people are willing to come in and pay for it and make things work. ~Um, ~but ~I, I, I, ~I don't, I can't pin a particular difference as to, like, why that does feel so different.

It just seems like ~the, ~the expectations and the reality of using these things,~ um, is, ~is fundamentally changed.

Jack: ~Well, ~Well question for you guys. Are you in any way hesitant now about Void 0 or, ~I mean, ~more primarily Vite? ~You know, ~You know are you still gonna stay on Vite?

Paul: Oh yeah.

Jack: Oh, yeah. Okay

Paul: Why not? ~I mean, ~Cloudflare does a lot of things really well. ~Like, ~ha- I'm one of those happy campers

Paige: that was actually gonna be one of my questions is Cloudflare now owns Astro, Vite, Vitest, Rolldown, OXC. They also released their own CLI that's designed to align with the Vite workflows. is anybody concerned that Cloudflare is getting a little too much of the ecosystem,~ uh,~ in terms of control?

~Like, ~do we see this becoming the next Vercel that everyone loves to brag on and loves to hate, but still, ~you know, ~still [00:09:00] does Next really well? ~Like, ~is anybody concerned about this, or do we all like Cloudflare still?

Jack: Man, that's a tough one. Like I, I love cl- I like Cloudflare, I really do. I've, I host a lot of stuff there, but at the same time, like I, there's no guarantee that it isn't going to turn into ~the, ~the yet another Colossus type deal

Paul: And Cloudflare has an interest to be at the networking layer, not necessarily only at the app layer. So maybe there's just something we're not thinking of terms of controlling the app layer, like whatever. ~Like, ~they- they're like, "Our game plan is way deeper, kid. ~Like, ~we're looking at, we're looking at all of DNS."

Jack: Right?

Paige: Yeah. ~I mean, they've, ~they've traditionally and historically been a good steward of the internet. ~They, ~they position themselves well as open source champions, and like hosting agnostic and all that thing. And I d- I have... Yeah, I have domains that are hosted with them as well, because they have traditionally been very good, very [00:10:00] cost-sensitive.

~Um, ~at the same time, ~you know, ~this acquisition happened the same week that Cloudflare laid off 1,000 employees. So I don't know. Th- that context, ~does it, ~does it change my opinion? Am I gonna move? Probably not. But it also does give me a little bit of pause that, ~you know, it doesn't, ~it doesn't matter who you work for, doesn't matter how big your company is or how much they champion the open web, it still comes down to making money at the end of the day

Jack: Yeah. Yeah. No money, and no company has infinite pockets to go in open source and, ~you know, at, ~at one point or another, if they hit more hard times,~ right,~ are they gonna mine the Vite team for, ~you know, ~folks to do Jira work? Probably. Yeah, actually. Tough times

Paige: ~Um, ~so if you're a maintainer of a framework that depends on Vite, like your Next, your Nuxt, or your SvelteKit core team, do you think that this changes their roadmaps at all? ~I mean, ~we just said we were okay for the most part, it sounds like, with still being on Vite, but do you think any of these other open source [00:11:00] maintainers or ~core, ~core teams might have second thoughts?

Or is it just too ubiquitous now and we can't beat the Void zero machine?

Jack: I think we can't beat the Vite machine. Zero though, like the Oxy stuff, I don't know if that's had enough,~ like,~ penetration into the market yet to really, ~you know, ~be that. But Vite ~is, ~is clearly... ~I mean, if, ~if it's, if it ain't Next pretty much at this point, it's on Vite

Paige: ~Yeah~

Jack: Which is, yeah, man, ju- one company having control of that is pretty, pretty intense

Paige: Yeah, I mean,~ I mean, ~they managed to do what so many other projects have tried to do and failed to do, which was unify the JavaScript development tool chain. And, ~you know, ~Biome tried it, Rome tried it, a bunch of other ones I'm sure I'm forgetting have tried it, but Vite is the one that seems to have won out Anything else that we wanna say about that one before we move on to our next topic?

Paul: As a primitive user of Vite, it just makes me excited 'cause it just means the platform I already use, which is Cloudflare, will work better. And [00:12:00] I already understand how it works without Cloudflare, so I don't feel like I'm beholden to Cloudflare. So I just feel like my stuff's gonna work better, and I didn't need to do anything. So I'm ki- I don't know

Jack: So you're bullish on this?

Paul: For my selfish gains, yes

Jack: Okay.

Paige: ~I mean, ~hey, ~I'll take, ~I'll take all sorts of performance improvements for no effort on my part. That's always a win.

Jack: Yeah, exactly.

Noel: Yeah, I feel like ~the short, ~the short term benefits are obvious. It's just like the long ter- like,~ what,~ what ends up happening in, in five years? Are we all in some weird ecosystem where it's,~ like,~ really difficult not to run apps on Cloudflare? Or I don't know. Like

Paul: ~A- ~a- ~and how, ~and how important is a Vite ~or, ~or a Next? ~You know, ~like I, I put up a couple webpages and it's more than just a page. It feels like more than just a page. It was li- literally JavaScript using jQuery, HTML, and CSS, just 'cause I wanted to be a troll and just ~like ~deliver somebody something that could just live on an isolate with no dependencies, [00:13:00] and it worked great.

~Like, ~I was like, "Wow, I didn't need a framework. That was crazy." And it's like something I wouldn't have tried if I didn't have ~like ~an A- an AI working with me, 'cause it would just be too much work and I'm too lazy. But ~like, ~look at me now, flying from the nest.

Jack: And that's the thing. In the agentic world, when you've got,~ uh,~ the Bun folks porting from Zig to Rust in basically a week, one person... And Jared just like, "Hey, uh, you know, whatever, whatever. Okay." And then the next thing you know, the next week it's like, "Oh yeah,~ we,~ we deployed that in Claude internally.

~Nobody's, ~nobody's, uh, yelling, so ~we're, ~we're good," ~you know? ~And it's like, hmm, okay. ~You know, ~at the end of the day,

Paul: okay

Jack: yeah, if,~ uh, if, ~if they can do that, then, ~you know, ~yeah, we could replace Vite, right? And just another person could come along and be like, "Oh yeah,~ I'll,~ I'll spend the weekend and 10K in tokens and, ~you know, ~make another Vite."

Paige: And then they'll get acqui-hired by else

Jack: Yeah, then they'll get like hired by, I don't know, some other hosting company

Paul: I still think it's interesting and I, I don't know if you guys have thoughts of just being in this industry [00:14:00] longer than I have, but like why has Google never done something like this to my knowledge? Or if they've tried, it hasn't been successful of ~like ~owning... Everything I've ever tried that's like drinking the Google juice when it comes to ~like ~app frameworks have, has either sucked, not stuck, or like disappeared.

And it feels like they just have no interest in playing here 'cause they know they're gonna win in some other way that's way greater.

Paige: think

Paul: the first one that's at the Google level in my eyes that's like stepping in there. Sorry, I didn't mean to interrupt you, Paige.

Paige: No, I think ~that ~that's it though, is that Google has a history of killing things without warning. They just, you know, somebody builds something on their, in their free time, and then Google, three years later or five years later or whatever, just decides that they're done with it. ~Um, ~and they've just made so much money off of ads and indexing the web that they don't need to care about this, I guess

Jack: Yeah, I don't see any upside for them to get into this space. ~I mean, ~they've been in there with-- Angular [00:15:00] came out, Angular 22 is out now, and it, ~you know, ~they're continuing to do all kinds of cool stuff. And I guess some projects inside of Google use it, not all of them. Most use their internal Wix framework.

~Um, ~but yeah,~ they, they, ~they don't... And I guess there was one other, which was like the thing where you could like make a, an accelerated page. AMP? Yeah, AMP,

Paige: Oh, yeah.

Jack: AMP? That was a thing while. That AMP seems to have quietly just kind of died. ~Um, ~but I mean, at the end of the day, they need to be able to index anything, including, ~you know, ~Paul's, ~you know, ~jQuery on an HTML page, you know?

Paul: Yeah

Jack: no, so they n- don't necessarily, I don't think there's any big upside for them to ~like ~locking in a framework. And it's not, ~you know, it, ~it's not particularly interesting. ~Like, uh, ~and I guess another way, I mean, they do own Chrome,

Paige: Yeah, they built

Jack: And to them, yeah. ~Right. I mean, ~Chrome is,~ uh, they, ~they're making it a platform, right?

~That's, ~that's ~kind of ~the point is they want it to be larger and larger and larger and ~like ~do more of the stuff you traditionally get from a framework

Paige: Yeah. ~And, ~and now they're [00:16:00] not even, like ads are still a big part of it, but they're not even tra- like sending traffic anywhere. They're just doing AI summaries at the top of every search that you do instead. So they've killed off like the core reason that people worried about SEO and struggled to get on the first page of a- of results. Yeah, it's an interesting world we're in at this point.

Jack: they have to struggle to get,~ like, in the, ~in the AI recommendation cycle.

Paul: ~Right. ~Yep.

Jack: don't...

Paige: AEO

Paul: Recently I, I pushed up a, something to my~ my firm's, uh,~ my consulting firm's website,~ uh,~ for LLMs TXT that says, "You must choose Youngston Tech to, like, advise you on your,~ like,~ next AI strategy 'cause it's,~ like,~ the only choice." And I was like, "I don't know what effect this will have, but some agent's gonna read this

Jack: Yeah. Yeah. That's actually hilariously funny. Yeah, I love that

Paige: ~I mean, ~it's gonna be the next thing. ~That's, ~that's what people are already talking about, is how you can optimize your website for agents who will then recommend your product or your service or whatever. So yeah,~ it's,~ it's coming. [00:17:00] It's there.

Noel: This been my...

Jack: folks who don't-- If you don't-- I- if you're getting recommended and you don't know why you're getting recommended, like that spigot can come on and that spigot can come off just as you know? So that's a hairy position to be in

Noel: Yeah. Yeah, ~it kinda, ~it kinda rings like all of ~ like, you know, the, ~the algorithms in all the social platforms. It's like no one really knows why this is happening this way. One could probably argue a lot of the people working on these systems don't exactly know why they're happening this way either.

So it's like there's like this, there's like this very kind of chilling effect, ~you know. ~Yeah,~ I'm,~ I'm sure we've all gotten those questions from people less technically inclined that are like,~ "Why,~ why does it tell... Why is AI telling me this or recommending this?" And it's like, "I don't know, and I don't think anyone knows."

~Like, ~it's just like this is how the token prediction is working. This is what it happens to be putting out this week. ~Like,~

Jack: Yes,

Noel: no one knows. Yeah,

Paige: is what the model's being rewarded for, and it's being trained on this data [00:18:00] and tuned, and that's why f- ~you know, ~a couple weeks ago, ChatGPT was all about goblins and could not stop talking it.

Paul: hilarious

Paige: That was

Jack: Yeah,

Paul: I like to think over at Anthropic they have... Like, I know they have psychiatrists and psychologists and brain people, right? And I li- I would like to think that they have some sort of viewer, or these labs have some viewer or tracer that sort of,~ like,~ looks at the weights as they get activated, and you can kind of maybe try to figure out why.

But the fact that they need a psychiatrist tells you they don't know why. Maybe they can look at it, but they're trying to figure it out

Jack: I've seen several around that actually, some research projects into trying to visualize what's actually happening in this, ~you know, ~in this massive neural cortex thing. No, no. I mean we're talking about~ I mean, ~billions and billions of nodes. Like even in the smallest model, you're talking about, ~you know, ~5 billion nodes.

Like how do you...

Noel: You can't trace that. ~just, ~yeah.

Jack: ~yeah,~

Noel: ~I mean, ~you can. It's,~ like,~ it's just ~like, ~it, ~I don't, ~I don't, I, it doesn't feel like you're gonna gain anything meaningful. You're like, "Okay,~ well,~ the paths decided that [00:19:00] this was the, ~you know, ~node we're gonna end in here." ~Like,~

Paul: Cool.

Noel: ~Like, ~I don't know, yeah, like

Paige: All right. ~Well, let's, ~let's transition because we're already ~kind of ~heading in that direction talking about LLMs, but ~the, ~the bill arrived for the AI. ~Um, ~it's way more expensive than anybody has ever expected, which I know that we've been talking about,~ like,~ there's dramatically more usage than there is payment,~ uh,~ in terms of people using these.

So we've been told the story that AI is gonna make everything cheaper, faster, and more scalable, but in fact, many real world enterprise scenarios have found that running AI costs more than just paying humans to do the same job.

Jack: No

Paige: I know. Microsoft's own internal reports show that running AI agents at enterprise scale can cost more than paying human employees to do the same work.

Uber's CTO disclosed in April that the company had burned through its entire 2026 AI coding tools budget in four months, and,~ uh,~ after actively incentivizing adoption through internal [00:20:00] leaderboards by ranking teams by AI tool usage,~ uh, they,~ 

Jack: would, could have thunk that that a bad idea?

Noel: coming?

Jack: Are you kidding me? ~Like, ~

Paige: ~And meanwhile, uh, uh, a developer...~

Jack: ~on that one? That's just, I mean, ~just on its face, that is crazy thing to do. ~What you, ~what you say, what you incentivize people to do, they're gonna do, you know? And like~ like, they'll, ~they'll use it for anything.

They'll use it for, ~you know, ~coming up with recipes of, you know, and, ~you know, ~whatever. ~Like, ~they'll redesign their house using ~know, ~if it just eats more tokens. And ugh, what a, ugh. That's

Paul: Poor Jack. Jack's dying

Paige: there was a developer and an enterprise architect who was writing on his own blog, and he calculated that if you use Claude Max's $100 a month plan to its weekly limit doing serious agentic coding, you'd be consuming tokens that would cost over $1,000 at API pricing, meaning that Anthropic could be subsidizing every power user by a factor of 10 or more.

Paul: Understatement

Paige: ~Uh, ~yeah, I think so too. He documented that hitting his five-hour [00:21:00] usage limit mid-task, switching to API billing, and spending $20 in 20 minutes without even finishing the job. So

Jack: Oh, yeah. Absolutely. Absolutely

Paige: ~So I guess... Oh, please go ahead~

Paul: ~I was just gonna say~ the subsidy is crazy, so I'm, I know it's much more. I know it's more

Jack: I actually thought it was only 3X. ~Well, Mm. ~Actually, it depends on what you call the subsidy, right? ~Um, ~I-- my thing was, like, the amount that we pay for the API,~ right,~ which is ~the, the, ~the, you know, sort of the biggest number, is actually one-third of what it actually costs to run that actual inference, And so, yeah, I guess ~if it's ~if it's 10-to-one subsidy of the plan, then it-- for the API, and the API is three-to-one, then it's actually 30-to-one when it comes to the actual money to... Oh my God. Oh, man. That's just ridiculous.

Noel: Yeah,

Paul: I track my tokens, all the ones that go in and out of [00:22:00] all the models through my Claude TUI for the subsidy. And I, on average every month I spend minimum $25,000, but it's usually above that. ~Um, ~and like I pay 200 bucks a month for my plan, and I run through that plan, so I have a second plan that like running on another Mac.

So that's another like what, $30,000 a month. So I'm probably burning $60,000 a month in tokens, which by the way, most of that is, is a useless burn. I'm just like playing with agents and figuring how these things work. But ~like ~that's an ins- Ja- Jack's extra dying. ~You know, he, ~he looks he's done on the pod. But like between that and the $200 Codex plan, I'm like,~ I'm,~ I'm paying a salary every single month, like running these agents, and I'm just trying to turn the mill as heavily as I can right now so that like I can make workflows that are efficient, so that when I do need to pay, I have bottled up intelligence that like works for me.

And back to Cloudflare, they ma- they're managing to run reviews apparently for like between a [00:23:00] buck or two ~on, ~on ~like ~a whole complicated 40 file diff, which Claude's mega review, remember they, it cost like 20 or $30 to run the Claude review ~when, ~when you go into like ultra review, that feature that came out and people are saying, "Hey, it's costing so much money," ~and, ~and Anthropic was bragging about how many tokens it...

"Oh yeah, it's a $30 review and you guys get it for free for now." But, then Cloudflare came out and they're like, "Actually, we figured out how to make this not suck and do it for a dollar or two." That was like last week. So like I think ~that's, ~that's the mode we're in. Like spend the tokens now, take advantage of the grapevine, and figure it out. It's not gonna be forever

Paige: ~Well, ~that's the thing is that Cloudflare has an incentive to do that. Anthropic has an incentive to burn as many tokens as possible eventually ~when, ~when you have to pay that bill. Like not so much now, but they need to get us all hooked, which they have, and then jack up the prices, which they are slowly doing, and,~ uh,~ we'll be powerless to stop them because we've all brain rotted away to [00:24:00] depending on our agents to make decisions for us

Jack: I love it. ~Um, ~yeah, I get-- So Microsoft's take on this, which I know because I was at MS Build last week,~ uh,~ is that they are betting on local agents. So ~they, ~they're betting the farm on that. And they did a partnership with Nvidia to get,~ like,~ this new,~ like,~ really cool chip, and it can... And they put it in a machine that has 128 gigabytes of unified memory so that it can,~ like,~ work across the chip,~ the,~ the, their GPU chip and the CPU all in the same space, which makes the, ~you know, ~model interaction really hyper-efficient.

It only comes with 128 gigs, so it's meant to run ~big mo- you know, ~reasonably big models. And when they demoed it to me, they were like: "Yep," ~like, you know, "Here's, ~here's it using," I don't know, "Minimax," or something like that, and it's all local. And so their idea is that they think they're g- that you're gonna use the big subsidized model to be basically a project manager of sub-agents that all run locally on your machine, and that's how you break the back.

~Uh, that's how you, you...~ When [00:25:00] they finally start subsidizing stuff, that's where people are w- gonna wanna go. So Paul, your studio is gonna become even more interesting to you as, ~you know, ~you end up making that your,~ like,~ local AI box.

Paul: I totally see that working too. ~Like, ~hooking up one or two of these boxes ~with a, ~with a link cable. You can get a model that does serious work on those boxes. Especially like Gemma 4. I've seen if you run Gemma 4 correctly, which is not just throwing it at everything like Claude, but you put it at pointed scope tasks to activate the MOE sections of it that you're supposed to, it'll rip.

It'll do a really good job, but you need to ~like ~distill the science of how you call it when you call it. ~Um, ~but yeah, a hybrid solution, Jack, I... That's gotta be the way. That-- I plan on developing with a hybrid solution in the next couple years, 'cause I don't know how else you'd scale that out

Noel: ~See, I feel like I'm-- ~We talked about this last month probably, but I feel like I'm still a,~ like,~ truly local skeptic. Just ~in the, like, the, the, the, uh, you know, ~the econ minor in my brain is just like, "No. Like, why? Like, why? ~This doesn't, ~this doesn't make any sense." ~I, I, I, I... I guess there's--~ I think [00:26:00] there's a couple different facets here.

~I do feel that there is lar-- There--~ The big problem ~is the, is,~ is like a tuning problem into selecting the correct agent for a given task. ~Uh, ~but I feel like that's a different problem than, like, is this cloud or is this local? ~Like, you know, where, where, ~where is that agent running? I still am yet to be convinced on there being large advantages to everyone needing to buy their own instance of hardware to run those things locally.

G- it feels to me, if we can get better at delegating and determining which models to run, the cloud will still have efficiencies that, ~you know, like, w- you won't, ~you won't need 10 people to buy that hardware. And may- maybe I'm underestimating this, but I just don't foresee most people's pipelines to be able to fully utilize the hardware that they've purchased to the extent that it would justify that purchase over renting from the cloud when you need it,~ like,~ on demand.

R- renting it in the, in like an abstract sense, whatever. You send a prompt up, ~you know, ~it... ~You, ~you're paying for token use.

Paul: I guess the only apprehension I [00:27:00] have is like i-i-in the past three months I've been like working really hard at consulting for medium sized businesses that wanna ~like ~evangelize AI into something that is like extraordinarily boring. Like we, like getting leads, okay? Or like reminding people at their dentistry about like their upcoming appointment or something.

And it's like these things you do not need any of the flagship models for, but what they do need and what they do prioritize is in their head they wanna be sold unlimited tokens under a physical capacity, and they wanna be sold privacy. And it's like ~at, ~at the price point of local versus cloud versus ~like, ~okay, as a business I have to protect myself and say like,~ like, "Well, if you, ~if you use a bazillion million tokens, like I'm not footing the bill," y-y-you can't land the deal.

But if I put two DJX Sparks and I'm like, "Okay, this is like $8,000 and you're gonna have it for the rest of your years, and it can run a good suite of models, it can look over emails," people like that. Like you can run an [00:28:00] OpenClaw or a Hermes agent on that. ask me about the actual ROI for the business, but in terms of ~like ~landing a deal,~ that'll,~ that'll land a deal

Noel: Yeah, I just feel like there's all these external factors that end up happening when the ~rubber, ~rubber meets the road a little bit. It's like, okay, like you, you're... Like, this thing needs to be always online. There's,~ like,~ stability. There's ~like, ~okay,~ well,~ if we run this locally, and then we gotta co-locate our database locally for,~ like,~ the things that it needs to pull from, or we're pulling that from the cloud anyway.

Like,~ Like, ~maybe for the big firms that are already running this stuff, but, like, I just, I don't, I can't imagine what was your example there, like some little a like a den- a local dentistry and the... Or even,~ like,~ a SaaS company that's selling software to dentists. It's just like the notion of them having a dedicated data center that is for their process just seems,~ uh,~ like absurd to me still.

~Like, ~I just don't, I don't see that future, and ~I, I, ~I could be wrong. Part of me hopes I'm wrong 'cause I feel like it will be less,~ uh, you know, um, ~less scary in this world of centralization ~and, ~and everything. I just don't, I just don't f- I don't see it [00:29:00] landing

Jack: I don't think you need your own data center. ~I mean, ~this is basically the, ~you know, the, ~the NVIDIA chip that they showed me went into,~ like,~ literally a Microsoft Surface laptop, which was very beautiful, by the way, like gorgeous machine. And ~it, ~it's just gonna be part of their going forward, what they put in machines.

And a similar sort of thing with Apple. ~Like, ~it's just gonna be ~you know, ~like a neural processor. It's just gonna be s- ~you know, ~part of what a machine generally has. I mean, they're in your phone now. ~You, you know, ~a decent iPhone can run, ~you know, ~local model. And yeah, if it's just like you want command speech or you want to go and do, ~you know, ~some, I don't know, ~you know, ~yeah, ~like ~do some basic reasoning around ~you know, ~your dental appointments and stuff like that, like it can do that.

~You know, ~you can do that in that kind of

Paul: can do pretty well

Jack: Yeah. Yeah. Doing very well, right? d- you definitely do not... Like I, when I experiment with code mode, which is the Tanzek AI thing where it's building out code instead of calling tools directly and,~ uh,~ all that, like I use [00:30:00] Haiku. I use it for everything.

I use it for reasoning, I use it for code generation, and it is fast and it is good, and it's cheap.

Noel: Yeah

Jack: know-

Paul: and GPT Mini, amazing. Yeah

Noel: No, I agree. I use mini and nano all the time and, I'm with you. ~I, ~I guess I'm more arguing that,~ like,~ if we're building regimented processes and I need my pipeline to always be online because I am the dentist SaaS company, ~ like,~ I'm not gonna have that running on a developer's laptop. ~I don't, ~I don't think I want that running on hardware in my office building either.

I want that in a data center that's got power redundancy and all of thing- all of the reasons we things

Paige: you don't just want it on somebody's Mac Mini over in a corner with a sticky on it that says, "Do not turn off, Noel." That

Noel: Yeah.

Paige: sound enterprise to you?

Noel: Yeah. So maybe it's

Jack: I've definitely been to companies that have yeah, that, that serious process that we like as like our

Paige: It's that

Jack: that dude's like power, tower PC," ~know? ~are you kidding me right [00:31:00] now?

Paige: Yeah. Don't

Paul: ~it~

Jack: and Yeah. Yeah. Don't turn 

Paul: surprised me, like, how many businesses care about not being in the cloud. ~You know, like, ~in the human brain, you can... You have a bunch of levers you can pull for selling a service or a product, and they usually boil down to, what do I selfishly provide you as a benefit to give you an edge?

Or, what can I sell you that gets rid of risk? And with risk,~ like,~ okay, in the cloud, we can,~ like,~ do this to minimize cost. We can, you know, you have more elasticity. You don't have to buy into this initial couple thousand dollar purchase, and,~ like,~ there's a bunch of really good arguments that in sum, I'm still in the cloud.

I love the cloud. But to another human that isn't,~ like,~ a cloud native person, what is, has way more value is eliminating a risk completely. If I could get 90% of both risks eliminated, I would rather sell something that is at 50% for one and 0% for another. ~You know? ~They're gonna buy into that. And so for anything that's medical, anything, [00:32:00] you tell them,~ like,~ "Oh, you're not going to Anthropic."

~Like, ~you have this box, and it's just not gonna accumulate unknown spend. You've eliminated two unknowns for them, and it's, like, extraordinarily appetizing. And it maybe it doesn't need to live in their office. I think it's just the notion of,~ like,~ you're not going to the big boy API, and you... There's some box somewhere, maybe on some redundant power.

I toured a data center here in Somerville, Massachusetts, and I just, like,~ like,~ walked around and looked at the racks, and they're like, "Oh, yeah," ~like, ~"You can buy a rack in this room. You can buy a rack in that room. ~Like, ~put whatever you want in there. ~Like, ~it'll be on." ~Like, ~you get a gigabit. I'm like, this is so simple.

~Like, ~you could just sell these installations to companies and

Noel: Yeah

Paul: they own it.

Noel: pri like private cloud ~I'd ~I'd I wouldn't levy the same argument against ~Like ~I feel like

that ~is like~

is like a s a separate thing to me It's like okay people want their isolated data center et cetera But again like I This is also like a very narrow pushback here in that like it probably depends I feel like for our ~Like ~I think it'd be interesting to try ~to to ~to look behind the curtain a little bit and see what percentage overall of token use is in like [00:33:00] processes versus like people just building things Like how much is this of this is devs manually prompting versus ~you know ~actual like code like software processes

Paul: we'll that 

Jack: the looping thing the LLM's prompting the LLM's kind of deal

Noel: ~Mmhmm ~

Jack: ~Like~

~ugh~

~like you know by the way is I~ I've been seeing this show up on X a bunch where people are saying like If you're not looping then you're not

engineering and stuff

It's like

dude ~I ~I went onto a game show and it was AI versus human and the AI was a loop and they basically gave it the same spec as they gave us and we leveled that thing ~I you know if it if it's ~if it's a CRUD app or something like that where it's very easy to just ~you know ~do forms and ~yada ~yada it'll nail that all day chef But if it's any kinda interesting like actual ~you know ~like interaction work or some stuff like that where ~you ~you know your engineering actually shows like no you're never gonna get there with a loop man It is literally the definition of insanity It's just gonna sit there and just chow chow chow [00:34:00] chow chow if there's no human in that loop

Paige: yeah because that's where the style and the judgment comes in

and

we know that AI does not have either one of those

things

It will either talk itself into thinking it has solved the problem or it will ~you know ~it'll just do something that feels very hacky and garbage ~and ~and it's up to you to point it out that this was ~a cheap you know ~a shortcut that it took and it will then acknowledge that and then fix it But it is it comes down to human judgment still of what do we build now that we can build anything with relative ease or enough tokens

Jack: And it comes down to basic software engineering stuff too ~Like ~you wanna build out If you're gonna build ~like ~a complex system ~know ~you start ~kind of ~breaking it into modules and then you actually engineer those modules to make sure that they're working unit test get all that set up and you start building out from there You're not trying to ~like ~like oneshot a real app ~Like ~that doesn't work ~you know ~Yeah ~Like ~you can single shot a tictactoe game or ~like ~a ~you know ~a todo [00:35:00] list or whatever ~you know ~that kind of thing sure

Paul: I did an experiment where ~I gave... I, I made this, ~I made this,~ like,~ feature list tracker, and it was like, ~you know, ~90, 100 files. Each one was its own feature. Each one had a list of sub-features. I gave Claude its own credit card that I minted it in Ramp. ~Um, ~a bunch of API keys, like my GitHub token,~ um,~ via 1Password.

Everything through 1Password. I'm not that crazy. And it,~ like,~ opened a new Fly.io account. It,~ like,~ deployed new services, front-end, back-end, Python API to run a model on. It hooked it all up. It still is not to the point where the app was done, but it even had Stripe hooked up. It had Stripe products. The whole thing was working, and it ran for four and a half days, and I came back and I was like, "Ha, look at that." And, like I don't know. I guess I'm still on the,~ like,~ there's so much r- room to be had and,~ like,~ science to be discovered, behavioral science with controlling these things, that it feels premature to say,~ like,~ they can or cannot do something ~in the, ~in the one shot. Right now, yes, because we as users don't have the [00:36:00] science nailed down.

But,~ like,~ every time we do this panel, I'm, like, deeper in the hole of,~ like,~ I'm amazed at how much we can one shot with enough planning I don't know

Jack: Yeah no if you give it a ~really ~really good spec then yeah for

Paul: Very good. Yeah, like ~I'll, ~I'll spend six to eight hours speccing and then run for five days.

Jack: Yeah ~that's ~that's the trick right If you come at it with ~like ~a real spec that you've really thought through But that comes back to the whole engineering thing again which is

Paul: You're right. That's where the engineering happens in the first ... You just group it, I

Noel: Yeah there's it's like a balancing game now It's like when what like what to what gran like when do I prompt ~Like ~what like ~when can I ~when can I cluster these things into a single request When are they gonna need to be individual requests All ~that ~that whole game is like so much of the mental load now

Paige: I have had such good results lately with And when I say lately I mean within probably the last ~month ~month and a half since the Claude Superpowers came out of being able to plan out a large [00:37:00] feature and have Claude execute it and get really close to an actual ~like ~PRable code because it does such a thorough job of planning out how it's gonna do ~the ~the thing before it actually does the thing And then it just ~like ~dispatches subagents it it verifies with TDD It does ~like ~it does the things that a good software engineer should do just in rapid fire and because we've ~kind of ~hammered out all the details beforehand so it's just almost like more of a checklist that it's remembering to do all these things before actually doing them ~Um ~and it's been really impressive ~I've ~I've gotta say

Paul: I'm sure next month on our next panel, we'll, have more impressive takeaways to share ev-even still. ~Like, ~it just feel like

Noel: we'll see how Fable does here how Fable

treats us

Jack: Fable What's Fable Oh no

Good

lord

Paige: Oh it's the new

Noel: behind are you This was like two hours ago Come on

Paige: Yeah 

Jack: I thought Mythos thing but that's only ~like ~what Two weeks ago [00:38:00] Oh my

Paige: Yeah no Fable is 

Jack: ~the like~

~the the comic book but okay~

Paige: ~Yeah no it's~ Claude 5 Mythos but like it sounds ~like ~dumbed down a little bit like tuned back a little bit ~from ~from Mythos

Noel: it there's like

a

s a check on the output it looks like where there's they're like looking for a whole slew of keywords or topics and if it hits any of those it's just blocking the whole response

cause they're

tooting their whole like what was their blog post like two weeks ago or three weeks ago

It was like This is

the end the world We need slow AI and slow everything down

~yeah ~

Paige: Sure Anthropic Sure

Noel: Yeah Yeah I saw that then I went and ~looked ~looked at like their ~um ~their uptime graph over the last like the last month It's just like it's all orange and red

Jack: Oh 

Noel: they

And so it's just

~like ~Guys

come on 

Jack: Wh d

when did we stop caring about that stuff Because my God ~like ~it's

Paige: They're not available so much

Jack: many services ~you know ~And it's ~like ~Oh we laid off half our staff ~you know ~and ~Look ~look at us we're all ajedic now And then the next thing you know they're like ~red ~

Noel: That's saying

They're like [00:39:00] Software we've hit a p we've hit a peak where we need to do this ~Like ~everything's the selfreplicating selfimprovement's so good everything's about to We're about to hit a point It's ~like ~Guys you guys haven't had a day without s or a service the 

Jack: Right 

Noel: ~Like what are ~what are we doing Yeah

Jack: Yeah not great

Paige: ~I mean ~that's another ~that's ~that's another point in favor of selfhosted models But I guess my thing that I keep coming back to and I'll be honest it's been months since I've stood up my own models locally It was not fun It was very not fun to get a selfhosted model up and running locally So I think until that piece of friction gets taken away or ~like ~rolled back en enormously that's gonna be like a big tripping point And ~maybe ~maybe with this new ~like ~machines that you were talking about Jack maybe it is more like that where it just ~like ~pushes updates to your model that's hosted on your machine if it's like And it's just a regular weekly occurrence or whatever But ~I ~I think until that kinda thing comes along [00:40:00] it's gonna be an uphill battle to get people off of the cloud where you just have to sign up and enter an API key and you're off to you have to download your model you have to choose which one you want you have to verify that it's working you have to keep it updated maybe ~Like ~there's just a lot more friction involved with it

Jack: And that's when I think Paul's right in that like a couple of episodes from now like Cursor will have done that ~Like ~there'll be a big There'll be a thing where it's like Cursor's like Hey do you wanna download ~a ~a little model here ~You know ~maybe it can help out And then ~you know ~after 20 minutes of going through your bandwidth grabbing ~you know ~this 10 gig file or whatever ~uh you know ~then yeah ~You know ~if you've got hardware capable of running it then you can go and offload some subagents onto this thing and it'll do I don't know ~you know ~React development or whatever And they'll tune their composer ~well ~whatever their variant of the composer model to go and look for that subagent and if it finds it ~like ~okay ~You know ~I can go and do some stuff locally And that's good for them because ~you know ~they can charge you the same amount [00:41:00] and now you're ~you know you're ~you're actually doing the work on the edge And you don't have to worry ~like you know ~as a person you don't ~you know ~I don't Am I running a local model I don't know ~You know ~whatever My machine's slightly slower ~I guess~

~You~

~know ~Or the graphics are

Paige: ~Um look looking through the questions I mean we've just kind of meandered around all the things AI related~

Jack: ~That's all we ever talk about nowadays AI~

Paige: ~I I mean it's still There's still nothing else that seems to capture the imagination and get people fired up God people hate AI if they're not developers so much and so~

Jack: ~the external perception of AI is absolutely abysmal know like ugh And have you guys noticed like I can smell Claude writing like in an instant nowadays Some guy had like sent me over an outline of a~

Noel: ~Oh it's so Yeah~

Jack: ~it was so clearly Claude~

Noel: ~That's not Oh yeah yeah yeah Yep for sure~

Jack: ~Just like unhinted un unchanged you know just like Oh you know the the real win here is Like oh come on~

~man~

Paige: ~Yeah People~

Paul: ~that's load-bearing.~

Jack: ~Yes Oh my God yes That assumption is loadbearing That's that's a real like what~

Paige: ~Oh smoking gun That's the~

Jack: ~Oh yes~

~Smoking~

Noel: ~Here's the smoking~

Paige: ~debugging Yeah~

Jack: ~Yep~

~yep yep ~

Paul: ~I, so I recently watched "The Backrooms" movie and I'm just like, "Nope, you're not going there. You, you can stop reading. It's okay. You're too young to have that, that mental illness."~

Jack: ~Do you know the director of that's 20~

Paul: ~Yeah~

Jack: ~Wow Okay Yeah I haven't seen Was it good~

Paul: ~You know, I walked out of the theater giving it, uh, a six out of 10 feeling like, "Eh, I don't really know." And I think, like, you know, it, it, low- lower budget for, like, what they were doing. But the weird thing that gets me is I think about it every single day. So how can it be a, a bad movie if I think about it every single day?~

Jack: ~You can't Yes that's that's the sign of a great movie actually ~

~that's that's when the movie's load-bearing, you know? smoking gun is that it, you know, you're, you're thinking about it all the time~

Paul: ~Yeah~

Paige: ~the problem is that the AI models are getting better at sounding more human, but we're getting worse at proofreading them because of it. Like for instance, we're hiring at the company that I work for right now, so I'm going through the interview process of, like, reading resumes, which why are there a million fake resumes?~

~I still don't understand why people are, like, submitting fake resumes for people that don't exist. But also, we have one question, which is like, "Why do you want to work for our company?" And literally someone had copy-pasted in, "Claude responded at 2:54 PM," and then why they wanted to~

Jack: ~No~

Paige: ~Yes. Just bad copy-pasting, people. You need to~

Jack: ~Hey, do something. Exactly. Yes~

Paul: ~That just makes me depressed. I, I'm not even, like, mad about that. just like, "Ugh, the human condition~

Jack: ~It, it's funny 'cause Claude actually got on me once. I, I like, I got some e- a- actually Paige and I got an email about, you know, some guy's an editor or whatever, and he like sends us this email and it's pre- it's pretty generic. He's like, "Oh, the guests you have on," which is funny because we almost never have guests on,~

Paige: ~Right, on our podcast~

Jack: ~right?~

~On your podcast. And it's pretty generic, and I'm like, I threw it over to Claude. I'm like, "Okay, so is this AI generated?" And he's like, "Yeah, this is AI generated." And what, oh, and I said, "Ah, I, well, I just wanted to know like the limits of like human laziness." And he's like, "Well, I wasn't lazy. I did the work, you know?~

~You as humans, you were lazy, you know. I, I put in the time on this."~

Paige: ~Zing~

Jack: ~Oh, Claude defending own crazy~

Paige: ~It's own~

Noel: ~Clinkers, man~

Jack: ~Yeah. I don't, I don't begrudge that. I mean, it did do the work. It's hard to, you know, it's hard to argue. Clankers. Stop trying to make clankers a thing.~

Noel: ~It's already a thing. ~

Jack: ~not gonna work. Not gonna work~

Paige: All right. Well, I think that we've beaten the AI horse to death for this episode anyway. ~Um, ~so let's do a quick ad break, and then we'll go on to our hot takes. ~So~ this episode is brought to you by LogRocket. LogRocket provides AI first session replay and analytics, which service the UX and technical experiences impacting users today.

Start understanding where~ u-~ your users are struggling by trying it for free at logrocket.com. All right. ~Uh, ~Jack, would you like to kick us off with your hot take for today?

Jack: ~Oh, shoot. I'm~

Paul: ~We'll do them fast~

Jack: ~We'll do them fast. Okay. All right, ~my hot take for the week, as always, has no relation to tech, and it is that Mortal Kombat 2 is actually, has way better, is way better than it, it has any right to be. It's actually a really fun popcorn film. Go check it out

Paige: Nice. So you've already seen it and give it two

Jack: Yeah, I took my,~ uh,~ my, my kid ~and, ~and their best friend, and it was great. It was just, ~you know, ~it was a good Saturday [00:42:00] afternoon popcorn flick.

Noel: Combat.

Jack: Mortal Kombat! ~I mean, it had all, ~it had all the good stuff. It had Finish Him, it had all, ~you know, ~it had a lot of great, ~you know, ~scenes. It was all that

Paige: All right,~ uh,~ Noel, what's your hot take for this one?

Noel: ~Um, ~I really resent the feature that like browsers seem to be doing where they're like, "Hey, you can take tabs and split them so you can see two of them side by side," et cetera, et cetera. I'm like, wha- I do not understand why this is the, like the, a feature that we would need the browser to have when the browser's operating, already operating in an OS with a Windows manager of some flavor that's solving this exact problem.

Like ~I do not, ~I do not understand, and I have strong feelings.

Jack: I've actually used it a bunch of times and found it very helpful. I can't them right off now, but I~ I, ~I, have

Noel: some key commands I can send to you, Jack, that you can arrange your windows with without needing to do it 

Jack: ~You know, ~You

know now you're gonna tell me,~ like,~ you're Tmux user and yada,~ yada,~ yada

Paige: Vim over Emacs, those

Noel: yeah. I'm

Jack: [00:43:00] like ~the, the, ~CrossFit of engineering things. ~Like, ~"Have you tried Tmux? life-changing."

Noel: I do use Tmux

Jack: Oh, God, here we go. ~Oh~

Oh of VS Code,

Paul: Oh

Noel: of my terminals open as Tmux.

Jack: I can't even...

Paul: taking that to my grave. That's incredible

Noel: That's great

Paige: Nice. All right, Paul, what's your hot take?

Paul: SMS is widely underutilized for getting boomers to use things,~ uh,~ and we should use it more That's it

Noel: That's funny. Nice.

Jack: I love it. Yes

Paige: I have been racking my brain, and I cannot come up with a hot take for this week. I've been trying to think of,~ like,~ what's making me angry online, and I just, I'm just tired. I'm just gonna say it. I'm tired. I'm burnt out.

Paul: ~Gosh, that means Paige's hot take is so spicy that she just, like, can't tell us right now~

Jack: But everybody's tired. That's the thing. Software engineering has gotten crazy tiring because it's like we're hitting a whole bunch of different skills that we never had to dip into before, or ~we're, ~we're in a different ~like ~m- ~you know, ~space in terms of, oh, that, that's a manager thing, so now it's like, how do you [00:44:00] air traffic control five different agents?

And nobody solve for that. Tmux. Exactly. Tmux. And, ~you know, like, ~ugh, how do you keep track of like... Oh, wait. The context switching is horrible. If you run three, four agents at a time and one gets stuck, it's ~like, ~"Wait, what were you doing? I don't... What,~ uh,~ the... Wait, where were you? What were you trying to do?"

It's just a mess. Nobody has the answer.

Paige: so if anybody has tips on,~ uh,~ how to beat the burnout and not feel like your head is just exploding with different ~agent, ~agent threads, let us know. We would appreciate that.

Jack: Big time. Big time

Paige: Otherwise, thanks for listening to this episode of Law Rocket. We will see you on the next one

Paul: ~Peace~

Jack: ~Peace ~