1018: Google fires Workspace CLI Creator

Google fires the engineer behind its Workspace CLI tool, OpenAI previews GPT-5.6 with three new model tiers, and Astro 7 lands with a full Rust rewrite. Plus: Coinbase cuts token costs with smarter routing, and more in this week’s Syntax Live Show Notes 00:00 Intro 00:34 Welcome to Syntax! 01:46 Google fires Workspace CLI Creator 12:30 GPT 5.6 Is Coming 19:59 GLM 5.2 Released 23:23 Astro 7 Rust Re-write 32:46 Cursor Announces iOS App 35:08 Scott’s Workflow: Herdr + Mosh + Termius + Tailscale 40:33 Coinbase Reduces AI Cost with Model Routing 44:22 wayfinder-router - Local AI Routing CLI 48:16 Token efficiency in models and harnesses Martin Woodward on X 52:34 performativeUI - react components for AI startups 54:31 Brought to you by Sentry.io 55:21 Reachy Mini Robot 01:02:21 FUTO Keyboard Swipe for Android 01:05:40 CSS Quake 01:07:35 HTML Invoker API is Baseline Available 01:13:17 Cloudflare Temporary Accounts for AI Agents Hit us up on Socials! Syntax: X Instagram Tiktok LinkedIn Threads Wes: X Instagram Tiktok LinkedIn Threads Scott: X Instagram Tiktok LinkedIn Threads Randy: X Instagram YouTube Threads

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  So I didn't get fired, but I quit soon after that. But it's just the it's the red tape of large corporations.

• 0:05Wes Bos?

  I think that them firing this guy was a big loss.

• 0:09Scott Tolinski?

  To me, that initiative is a positive trait and like you should harness that rather than being like, all right, you're out.

• 0:17Wes Bos?

  I think a lot of them are just like, let's get as many paying customers in the door right now and we'll worry about the profitability in the in the coming months as we IPO.

• 0:34Speaker 00

  Welcome to Syntax. Welcome to Syntax Weekly, your weekly live show here with the Syntax crew. This week, we're going to be talking about how Google fired the Workspace CLI creator. Also, GPT 5.6 is coming. We're going to tell you all about it. Astro 7 was rewritten in Rust. And we've got a lot of other fun stuff to share with you as well. Today, I am joined by Scott and Wes. Scott, how are you doing today?

• 1:01Scott Tolinski?

  Me? I'm doing good, man. Doing good. Yeah, just chilling, big chilling. In fact, Wes and I were just talking about in the summer, the AC absolutely freezes the level, first level of the house and the second level gets like boiling hot. And I'm literally chilling, literally freezing.

• 1:21Speaker 00

  Same thing. Yeah. What about you, Wes? How you doing?

• 1:23Wes Bos?

  Yeah, freezing out as well, but stoked to be here. So to talk about all the crazy stuff that's been going on last week. Lots to talk about.

• 1:31Speaker 00

  Definitely. So if you're watching us live, throw a message in the chat. Say hello. Also, if you have more news to share that we don't cover, throw it in the chat. We'll be talking about that at the end of the episode. And without further ado, I think we should jump right in. Wes, what have you got for us?

• 1:47Wes Bos?

  Yes, this is kind of a crazy one. So Justin Ponell, who was a creator of the Google Workspace CLI. So a couple of months ago.

• 1:58Wes Bos?

  Google seemingly put out a CLI to interact with Google Workspace, meaning that you can interact with your calendar, your email, you know, all of the Google Workspace products. And at the time when it had launched, like, everybody was like, finally, like, this is exactly what I want. I want a CLI that I can give to my agents to be able to interact with all of my Google products. And I posted about it. I got over a million on my own, my own post alone. Like, like, that's how big this was. It was just like, everybody was freaking out about it. And they released it. They released it on the, like, official Google Workspace GitHub. And when I had posted it, everybody's like, this is not officially supported.

• 2:42Wes Bos?

  And it was really weird. Like, like, everyone was like, this replying, it's not officially supported, but it's, it's like, on the Google Workspace thing. Adios, Manny posted about it. Like, he's like, announcing the CLI from Google. And, and like, he was working in an AI role at the time at Google. So like, it, it seemed like it was official and everybody was really happy with it. I used it, worked really well. However, it seems like the guy, very shortly after, Justin, who built this thing, was canned and, and Google had fired him. So he posted on Twitter, "Two months ago, I was fired by Google for creating the Google Workspace CLI. It went viral, et cetera, et cetera.

• 3:24Wes Bos?

  I think the cause was that Workspace and certain leaders in projects were afraid of being disrupted. The fear wasn't specific to my CLI. It was a broader fear in what agents meant for Workspace. Either way, the irony of my termination, the announcement at Google Cloud, next two days before I was fired in the official Workspace." Oh, sorry, am I not sharing my screen here? I am.

• 3:48Wes Bos?

  I am the worst live streamer ever.

• 3:50Speaker 00

  Professional live streamers out here.

• 3:52Wes Bos?

  Professional live streamer. Here we go. Here's the post.

• 3:57Wes Bos?

  So he basically said like they announced their own version of the CLI at the time. And it's kind of interesting. Like everybody was freaking out being like, oh, like, why would Google do this? Like this guy, he was like Dev Ralph for Workspace. And he saw that people had a need of like, I want to be able to interact with it. There were some third party things out there at the time, but there was no, nothing that was like official, basically the entire Google SDK, but in a CLI. So he, he obviously made one up and, and released it to everybody. And then he got fired for that, which everybody's like, why would you do this? He obviously saw a, a need for this and, and put it out there.

• 4:39Wes Bos?

  And I, I think like, there's a kind of like two ways that like, why this happened, right? Like there's one, which is like, he just said, like the internally at Google, there's like, just like smaller businesses and they're all fighting for their thing. And, and if somebody comes around and makes a product that wasn't officially blessed, or like changes how their job, you know, like at big companies, people internally are always just fighting for whatever it is that they want. And they don't necessarily care about like the end product, which is the, the consumer. I certainly think that happens inside of Google. And then the other aspect people are saying is like, you just can't make stuff and release it under officially under Google. And like, like there's, there's processes and apparently this wasn't actually approved.

• 5:25Wes Bos?

  So maybe just by not following that is, is one of the reasons why he was canned as well. I don't know. What do you guys think?

• 5:33Speaker 00

  Yeah, I think big company politics, like this happens.

• 5:38Speaker 00

  This, I mean, I did this sort of, I can give a personal story because one of the first companies I worked about, worked at was a Fortune 500 company, really big. And I liked to move fast. I just liked to build things. And there were multiple little projects. So I, and we had a lot of red tape, right? Every project you create, you had to create a security assessment. It had to be approved, like, what dependencies am I using? What network areas is this thing gonna touch? And where is it gonna live? And so I had to fill all of that out for every single thing. But I, working there long enough, I figured out a path of least resistance to like, get my stuff out really fast. And I released something that like, half of the company started using. But because there wasn't an official project around it, I got like, reprimanded.

• 6:22Speaker 00

  I got brought into the boss's office and they're like, I need to know everything you're working on. So I didn't get fired, but I quit soon after that. But it's, it's just the, it's the red tape of large corporations. Like, they wanna be in control of every release and every product. So it's hard to move fast and it's hard to release new things when you're at such a large organization.

• 6:42Scott Tolinski?

  Yeah, yeah, it is, it is. And as Paulo says here, someone also said that the GitHub account was not actually from Google and he just created a Google Workspace account with the Google logo. And this is what I appear to have happened, appear to have seen happen as well. Because I think that is a big part of this too, is that I remember people asking very specifically, is this account actually a Google account? And then the way that people determined that it was, was by seeing that the guy who created the repo worked at Google, not that it was associated with Google directly.

• 7:24Scott Tolinski?

  So I, I, I think, I think both sides could have done this better. Like, instead of just tanning the guy, you'd think that they could work with him to empower his initiative. And like, Google used to be about that.

• 7:38Wes Bos?

  There's 57 repositories under this Google Workspace thing. Which, it tends to happen, right? Like, somebody who works, especially if you work in DevRel, you don't wanna like, post all of your samples and examples and whatever. So you make your own GitHub organization. I think Mozilla has something like 900 different GitHub, like, like, name spaces, right?

• 8:00Scott Tolinski?

  Yeah.

• 8:01Wes Bos?

  But like, they should have caught that.

• 8:03Scott Tolinski?

  But still, it's like, it's like, to me, that initiative is a positive trait. And like, you should harness that rather than being like... All right, you're out.

• 8:13Speaker 00

  Exactly, yeah.

• 8:15Scott Tolinski?

  Yeah, I think there's a proper way to reprimand and, you know, take that positive part.

• 8:20Wes Bos?

  You want people to be like moving fast, breaking things. There's obviously, you have to follow some sort of procedure to make sure you're not like putting out stuff that is insecure. But some of the best products are just gonna come out of people like that. Google has this whole 20% time, meaning that you can take 20% of your time and build whatever you want, because those are some of the best ideas that come out of here. And now they're being, now they're giving them a big bonk hammer, being like, well, don't, don't actually release anything, you know? I, I think that them firing this guy was a big loss. Like, somebody did something that was, I think like, he probably should have gotten trouble for what he did. But I don't think you should have fired him. You should have said, all

• 9:03Wes Bos?

  right, well, what else do people need? You know, like, how, how can we use you to move faster in our race to win this AI war?

• 9:12Speaker 00

  Yeah, I, I think it, there, there might be stuff we don't know, right? So like, maybe there was something he was doing that maybe this was the last straw or something like that. But at the same time, we, I don't know if this, if this is linked or not, but we know of other people working at Google that were like slightly associated with the project that also got let go. So that kind of leads us to believe that it's likely that this is the thing. I don't know, there, there should have been more process. I, I feel like there's, there's always like such like really smart engineers, people that have been there for years at all these companies, big companies, that end up getting let go for the silliest reasons or get, get caught up in layoffs. And it just goes to show that like, companies don't care about their employees, for the most, well, big

• 9:57Speaker 00

  companies anyways. They, they will fire you at the, at the drop of a pin.

• 10:01Wes Bos?

  Yeah, sometimes you also see these, these firings as well. And People are all up in arms, but you kind of, you kind of hear, I don't know if this is the case, but sometimes you hear internally like, yeah, that guy kind of sucked for the long.

• 10:15Scott Tolinski?

  Or it's like you get the one side of the story.

• 10:18Wes Bos?

  Yeah.

• 10:18Scott Tolinski?

  Yeah, you, you get the, the one side presented and then you didn't get the other side about what

• 10:22Wes Bos?

  Yeah, like what if there, there may have been some huge security vulnerabilities in this thing. And then somebody submitted to a bug bounty, you know, like there, there probably was a bunch of headaches that happened because of this thing. And that was probably very frustrating, you know, like people bring it to the lawyers. People like, the crazy thing about the CLI is that you were just like giving it full. Well, you could, you could, you could have like fine grain control over what it had access to. But a lot of people were just being like, sure, full OAuth access to every single scope of every single product I have via in a CLI. And then people were just throwing that into their open claw.

• 11:06Wes Bos?

  And can you imagine the like headaches of like this, this, this is a huge security issue as well. Just having that wide open on a CLI where other tools are, are running at the same time. So there may have been some like, hey, you can't just like go nuts on this type of thing. We have to vet the security and maybe even, I don't know how the new workspace CLI works and how like it's agent integration, but there may be a couple more safeguards in place.

• 11:37Speaker 00

  Yeah, I'm guessing there was some internal process that he didn't go through. And that's basically what got him canned because there's very likely like every project has to go through PR, has to go through security and maybe one of those things was skipped. And that's the

• 11:51Wes Bos?

  I think all of it was skipped.

• 11:52Speaker 00

  Yeah. Yeah.

• 11:57Wes Bos?

  All right, let's get on to the next one. What's our next story here?

• 12:00Scott Tolinski?

  Yeah, next story. So for folks just joining us, we, we do this live stream every single week where we bring a news item, something that's interesting happening in the web world. And if you have things that you want us to share, go ahead and post them in the chat. We're on YouTube, Twitch and X, the everything app where in all those places.

• 12:24Scott Tolinski?

  So yeah.

• 12:25Speaker 00

  What do you got?

• 12:26Scott Tolinski?

  Yes, me. What do I got? Let me share my screen real quick. So you may have seen on June 26th, so right at the end of the week last week, previewing GPT 5.6 Soul, a next generation model, which is essentially three new models from OpenAI, the 5.6 series of models. And once again, OpenAI has chosen to add more, a new naming structure for these things. And I feel like they just eventually have to do like naming bankruptcy and just like establish like a whole new structure because now we have

• 13:10Scott Tolinski?

  GPT 5.6 and then three new variations of that where it's Soul, Terra and Luna. And if you don't know why those are named that way, it's because a Soul Sun is the biggest, then Terra, you know, Earth is the next biggest. And then Moon, which is the next biggest, is the smallest of the the three here. And so how good are these things? Well, okay, so the way they're they're basically describing it is that this Soul is the best model they've ever released. It's the number one. Soul Ultra has the biggest, highest chart. There's never been a chart higher than Soul Ultra.

• 13:53Scott Tolinski?

  Soul Ultra is kind of a badass name though, for being honest. And yeah, they have it above Mythos. You know, okay, whatever. Either way, it's a big ass model. Next, next biggest one is Terra, which is more competitive with GPT 5.5 while being two times cheaper. And then you have Luna, which is weaker than 5.5, but cheap and fast, right? They also have Terra as being about the same as Fable, which I don't know. In my testing, Fable was significantly better than GPT 5.5 personally. But I don't know. This is their own their own numbers. Who cares? I think it'll be more interesting to see.

• 14:36Scott Tolinski?

  Different benchmarks over time. So yes, we have three new models from OpenAI.

• 14:42Wes Bos?

  And we can use them, can we?

• 14:44Scott Tolinski?

  No, hell no, you can't use them because you're a poor and you're a nobody. And you don't, you're not friends, you're not close, personal friends with the, the OpenAI folks. So therefore you get to use it whenever they feel like giving you some table scraps. But yeah, so word on, word on the street is yeah, they're, they're being able to be used by certain partners right now, except for over time, they'll have more information about when you, normal folks can use them. So you can't use them yet, but let's actually talk about actually some of the more interesting things about Soul Beyond it just being like a new, bigger model. They have Ultra Mode. Actually, a lot of these things, I will say,

• 15:25Scott Tolinski?

  are kind of similar to what the latest Claude models have been doing where in they, there's Ultra Mode, Max Reasoning, but it, it's inherently better at using subagents for complex work, spawning subagents for work. So instead of the harness being in charge of like, I, I guess, what, what was it? Opus, what was it? 4.8 that did subagents really well with Ultra thing? Not even Ultra Think. God, what was that, that new mode in Claude?

• 16:00Wes Bos?

  Um, yeah, it was the one that Bun was built with, right?

• 16:03Scott Tolinski?

  Yeah. And it's not Ultra Think, it's something. Um, God, I don't know why I'm blanking on this.

• 16:08Wes Bos?

  Some kind of loop, right?

• 16:09Scott Tolinski?

  What?

• 16:09Wes Bos?

  Some kind of loop?

• 16:10Scott Tolinski?

  No, no, no, no, no. Um, it's the thinking, I have Claude open right now. No, I accidentally just did some, okay.

• 16:18Wes Bos?

  It was like Ultra Code in the chat, Ultra Code.

• 16:20Scott Tolinski?

  Yes, Ultra Code, yes. So it's spawning subagents to do complex work. Obviously, it says that there is improvements in coding, biological stuff, cybering, beats 5.5 using fewer tokens, AKA it's a bigger, newer model. So, yeah.

• 16:41Scott Tolinski?

  I wonder if it's gonna be good at design. We'll see. Yeah, I have no idea.

• 16:44Wes Bos?

  Do you think the United States government will allow anybody else to use it? Do you think as a Canadian, I will ever get access to this type of thing?

• 16:53Scott Tolinski?

  Well, I mean, that's what they, that's one of the reasons why they're doing this slow rollout. It's the same, it's the same line. It's that the US government

• 17:02Scott Tolinski?

  wants to limit it. It's too good at cybering. It's the best model at cybering of all time.

• 17:08Speaker 00

  Yeah, I read a Washington Post article this morning that mentioned they're giving early access to the UK government.

• 17:14Scott Tolinski?

  Interesting.

• 17:15Speaker 00

  Yeah, so some allies of the US are getting access, not just, not just the US. I don't know if Canadian is that close of an ally to.

• 17:22Wes Bos?

  Oh, we're a huge ally. Come on, hook it up. OpenAI, if you're listening. Give it to Canada, give it to me. I will only use it for good.

• 17:30Scott Tolinski?

  Yes, felt interfaces. Could we also say that this is maybe one of the hardest to read charts of all time? Like, why, one, why are all the colors so incredibly similar on this?

• 17:44Speaker 00

  Oh, true, yes.

• 17:45Scott Tolinski?

  It's like shades of blue that are all like, look, bro, you can't even tell they're different colors. Like, okay, I don't know. I saw this. I was like, that's insane. This is an insane chart. So shout out to OpenAI.

• 18:01Wes Bos?

  Can I just say I'm so sick of all of these, how confusing all of this stuff is? Like, you said, like, there's eight different new names and there's, there's 11 different models and all the different features and whatever. Like, we need to get to a spot where you can just ask it to do what you want and it will figure out how to best route that and how to get you the best possible output. Because like, we'll see here every week being like, yo, they released a new 6.9.2 and, no, 6.7 is now better than X, Y, and Z. And then you're just, I don't know. It's, it's so frustrating because it's like, your day to day, it doesn't.

• 18:44Wes Bos?

  It's so hard to see if it's actually better at what you want and you having to choose which model you're doing for each one. We got to get to like auto routing for these things much faster.

• 18:55Scott Tolinski?

  Yeah. Are you ready for Codex GPT 5.6 Soul Ultra? Like how many descriptors can you throw on these things? They're really reaching the Sony WX M1003 naming marketing level here. Can we, uh, somebody from Apple get in there and shake things up on how to name a ding product?

• 19:20Scott Tolinski?

  Yeah. Well, that's all I really got. I haven't gotten to use it yet, which is sad for me because it shows just how much of a nobody I am. I don't have access to this model just yet. So, no, I, I'll be stoked to try it. I don't, I don't love the GPT five models because I do a lot of creative work and I feel like they're not as good as creative work. But you know what? Anytime a new model is released, it's always exciting to see and try and do all that stuff. And for exact kind of technical work, the GPT stuff is good, very good. So when I get my hands on Soul, you best believe I'm going to grip it and rip it.

• 19:58Speaker 00

  Nice. Yeah, I don't have any prepared news for this, but GLM 5.2 was released, I guess a week ago or so. This is an open weight model. And that's also exciting because a lot of the benchmarks they're showing, like it's on par with some of the frontier models like Opus 4.6 and GPT 5.5.

• 20:18Scott Tolinski?

  TJ.

• 20:19Speaker 00

  It's, yes.

• 20:20Scott Tolinski?

  Can you explain what open weight model means for the audience?

• 20:24Speaker 00

  Absolutely. So that means that the weights of the neural network are released as a file that you can download and run on your computer. So these models, like we just talked about GPT 5.6, the only way we can access them is via the OpenAI API, right? We just get remote access to them. Whereas open weight models, you can run on your own hardware or run on cloud, uh, cloud hardware. And that way you're not dependent fully on a provider like OpenAI or Anthropic. GLM 5.2 is pretty big. I think it's like a one trillion parameter model. And it, um, the 4-bit quantization is like close to 300 gigabytes. So you, you need a really, really beefy

• 21:09Speaker 00

  machine if you want to run it locally, but you could run it in the cloud. Like if you rent some GPUs. Uh, but the reason this is interesting is because as models get more expensive in terms of trying to interact with Anthropic or OpenAI, this is something that's potentially cheaper and you potentially could run on your own hardware for like privacy reasons, but also just like local compute reasons.

• 21:30Wes Bos?

  Yeah.

• 21:31Speaker 00

  So we're getting to the point where these open weight models are getting really good.

• 21:35Wes Bos?

  And, and you can run these models, like, like you said, you can, you can like rent your own compute or you can just go to like something like OpenRouter. And I think they have, I'm just looking at it right now, probably 15 different providers, um, for these models. So I know people are like, oh, I'm not trusting a Chinese model. Um, well, you just go to an American company that is already running it and it's still is going to be significantly cheaper than a lot of these other ones. So I need to, or you can even use it in like in Cursor or, um, uh, VS Code.

• 22:12Scott Tolinski?

  Yeah. I look forward to the day where model routing is, um, smooth enough that again, you can use smaller or open models or whatever you can run locally here and there in a way that you have confidence that the right task is going to be able to be accomplished quickly and easily by a cheaper, faster, easier to run model, whatever. And things are routed in that way automatically and successfully without you having to do too much.

• 22:39Wes Bos?

  Cause I don't like doing too much. I'll let CJ go through his story, but I have a story about Coinbase doing exactly that. So that's coming up.

• 22:48Speaker 00

  And I have a project I'll share that lets you do it locally too. So.

• 22:52Wes Bos?

  All right.

• 22:52Speaker 00

  Let's, let's keep it moving. Um, if you're just joining us, this is Syntax Weekly, uh, our weekly Monday livestream where we all chat about the news, new libraries, new tech, all the stuff happening in, in the world that we live in. Um, and, uh, we talked about, um, how Google fired.

• 23:12Speaker 00

  Someone for creating the Workspace CLI. We just talked about the release of 5.6 and I am here to chat about the release of Astro version 7. So, if you're not familiar, Astro version 7, oh Astro itself is a framework for building websites. They call themselves the web framework for content-driven sites. You can think like, marketing sites, blogs, e-commerce, that kind of thing. And it's really good. And version 7 got even better. So you can check out their V7 blog post. And it's all about Rust. Now, this is a common theme in the world of web development, like so many developer tools are being rewritten in Rust. Bun got a Rust rewrite.

• 23:54Speaker 00

  Rolldown is the Veet bundler and it's written in Rust. Rust, Rust, Rust, everywhere. But Astro 7 is here. And the thing about Astro is you can actually use it with multiple frameworks. So you can have React components, you can have Svelte components, you can have Vue components or Solid components. But typically, the top-level pages, you might use Astro components. And so this is their own custom syntax for building out components. And that's the thing that was rewritten in Rust. Essentially, the compiler for .Astro components. Rewritten in Rust, builds are anywhere from 15 to 61% faster in their benchmarks. And the other cool thing about all of this is you combine that with the fact that they

• 24:38Speaker 00

  also are using Veet under the hood. So Veet 8 with Rolldown, which means if you are using React components or Svelte components or whatever else, those are going to compile really fast or bundle really fast because they're using Rolldown and now your Astro components are going to bundle really fast. Basically, it's a whole lot faster. There's a lot of other improvements as well, including advanced routing. Let's talk about...

• 25:05Speaker 00

  Yeah, in terms of the Rust compiler, one of the things that their previous compiler did was it kind of behaved like a browser in that you could write non-valid HTML and it would still work. But with their new Rust-based compiler, it doesn't auto-correct your HTML. So essentially, if you don't have a closing tag or other weird issues that you didn't, maybe you didn't even know you did wrong, it's actually going to yell at you and it won't compile. So your code will be cleaner and more compatible across the web going forward. And it handles GSX-style strictness and white space a whole lot better. It is interesting to note that the previous Astro component compiler was actually written

• 25:49Speaker 00

  in Go, so another native language, but they chose to rewrite it with Rust, which is fascinating.

• 25:54Wes Bos?

  I wonder why. Does it say?

• 25:57Speaker 00

  Yeah, so it allowed us to ship native binaries for supported platforms with a WASM fallback for environments that need it. So I guess that might have been harder to do with Go.

• 26:06Wes Bos?

  Okay.

• 26:07Speaker 00

  I think you could do something like that with Go, but maybe there's better WASM support, like Rust compiler in script-in WASM support kind of thing.

• 26:16Wes Bos?

  Oh, cool. Man.

• 26:18Scott Tolinski?

  Yeah.

• 26:18Wes Bos?

  That's, the tooling is unbelievable in the last couple of years, especially in the last six months. Everybody's like, well, it's not a big deal now to be able to rewrite absolutely everything.

• 26:31Scott Tolinski?

  I think part of that, too, is that AI is really good at Rust.

• 26:36Wes Bos?

  Yeah.

• 26:37Scott Tolinski?

  And honestly, one, I want to, some comments were talking, saying that in our last video where we did our tier listing, that the Rust compiler is not slow. Bro, the Rust compiler is slow. But actually, that's not a problem with AI because you're like, just having an agent go on and work on stuff. I'm not the one sitting and waiting for the dang compiler to tie its shoes and do whatever it's got to do. I think that's pretty nice, working with Rust and AI compared to me handwriting it before. Also, I'm not good at Rust, so that's probably one of the reasons why I like using AI with it.

• 27:15Speaker 00

  I have multiple side projects that I spun up with Rust because AI said it would be faster. Like, I was running into performance issues and was like, let's rewrite it in Rust. I was like, okay. I haven't looked at the code anyways.

• 27:25Scott Tolinski?

  Yeah. Okay, sure, whatever.

• 27:29Speaker 00

  Oh, that's great. Some other fun stuff is they also replaced their Markdown and MDX pipeline. Basically, it's like full, fully new internals. And it's all really, really fast. So there's this library called Soteri, and it's a Rust-powered processor for Markdown and MDX.

• 27:49Speaker 00

  So a lot of times when you have these content-based websites, you might have your blog posts written in Markdown, and that needs to be processed, turned into HTML potentially. With MDX, you can have like component imports or like one-time scripts that eventually get bundled and compiled into the final HTML output. And so having that as a native pipeline is really great as well. And the thing that I'm excited about is advanced routing. Check this out. So they made it so that you can have a file called fetch.txt. And inside of there, you can do absolutely anything with any of the incoming requests for your Astro app. And so this pattern of fetch with access to the request, it was kind of popularized by Cloudflare workers.

• 28:33Speaker 00

  If you've ever written a Cloudflare worker from scratch, that's how you do it. You get access to the raw request, and then you can do anything. You can look at the URL, you can look at the body, and decide what to do with the response accordingly. And so they support that syntax directly. But that also means because they support that, it supports Hono, right? Because Hono is a way of building APIs, and it's basically built on top of this concept where everything just receives the request and can operate on it. So instead of writing everything manually, you can now bring in Hono and get access to all of the middleware's available in Hono. And if you want to have more structured API routes, you can build those out with Hono as well. So this is great.

• 29:13Speaker 00

  Yeah, in this example here, they added the internationalization, Hono middleware, auth middleware, actions middle, like all of these things. And all they had to do was export the Hono app. So right, so like, if you look at the raw code, you're, you're writing a fetch that gets access to the request and does whatever else. But because an exported Hono app has a fetch method on it, it can do all of those things as well.

• 29:35Wes Bos?

  Oh, man.

• 29:36Speaker 00

  So, super exciting.

• 29:37Wes Bos?

  That's great.

• 29:38Scott Tolinski?

  Astro rules.

• 29:39Speaker 00

  Yeah.

• 29:40Wes Bos?

  I had never gotten into Astro and I, it's, it's really weird because I absolutely love everything that Astro does. And I love like the, the methodology and they were the first on like the Islands architecture and all of that, but I just had never gotten into opting into their own custom like language, you know, like the .astro files. They're always just a bit too weird for me and I, I never had made the jump to it.

• 30:07Speaker 00

  Yeah, I built a couple of content sites with it, but they just, they've added so many things like they have like a database layer. Now they have the API layer, like you could build full, like full stack apps with this. And it, it, it sort of initially wasn't meant for that, right? It's meant for content based websites, but they have all the stuff built in that you potentially could. And I have one more thing to share in the announcement and this is an AI enhancement, which is they include a background dev server. Now, every single CLI needs to do this because I don't know how many times I've been building something with AI and it tries to start the dev server or start multiple dev servers or like it can't find the port or whatever else.

• 30:48Speaker 00

  So now, if you add dash dash background, it has a nice little output that says what the process ID is and lets AI know, hey, there's, we already ran this in the background and we'll keep it, keep it running in the background for the AI.

• 31:01Wes Bos?

  Yes. Next, I added this and I built my own Vee plugin exactly for this because it's so frustrating when it spins up 10,000 individual servers for it. So I'm a big fan of this.

• 31:16Scott Tolinski?

  You know what I've been doing now is I just set an explicit port on every single project. That way Vee doesn't do the incrementing thing or the 5174.

• 31:25Wes Bos?

  Yeah, but, but even then, like I, I set a, a specific port for absolutely everything, but then it like kills it because it like, I'm running it, but then it thinks that like it should kill it and restart it itself. And it's, it's so frustrating to, to do that. So the, the strict port.

• 31:46Wes Bos?

  Is good, but the strict port doesn't like explicitly kill it if something is on it. So what this does is it it puts something in the in the CLI where you can explicitly kill it and then it's then it's good.

• 32:02Scott Tolinski?

  Yeah, that rips.

• 32:03Speaker 00

  Yeah, Astro team doing great things and just like moving with the times, right? Like they're they're fast, they're using Rust, they're taking into account using AI to develop. They're doing it right over at Astro.

• 32:17Scott Tolinski?

  That's right. That's right, CJ.

• 32:21Scott Tolinski?

  I love this. Yeah, shout out to the Astro team.

• 32:24Speaker 00

  I think Wes has something if we want to chat about that.

• 32:27Wes Bos?

  I got I got 10 million things here. Let me let me show you here.

• 32:32Speaker 00

  And if you're just joining us, this is Syntax Weekly. We're talking about news, tech, things we're interested in. If you have things that you want us to talk about, throw them in the chat. We'll talk about them at near the end of the show. But yeah, Wes, what do you got?

• 32:45Wes Bos?

  So Cursor just announced their iOS app. They they teased something this morning and I was hoping it was Cursor Composer 3, which is their model that they've been working on. Cursor Composer 2.5 is actually pretty good, but apparently their Composer 3 is going to be like at the level of like an OPUS or an OpenAI coding model. And that's the way it's going right now, you know, like Cursor has got to come out with their own equivalent model in order to sort of survive these things. And I know like Devin is working on their their own model as well. All of these tools that are using other people's models are like are like kind of suffering, right? Like they everybody's got to come out with their own very good model.

• 33:30Wes Bos?

  But that wasn't what they released. They released Cursor for iOS, which will allow you to either like work with your cloud agents, which you could always use the website for that. So Cursor has this idea of cloud agents where instead of running it on your computer, you just you run it in their cloud and they give you like a they they give you just a Linux VM that has a browser attached to it. Well, and and you can just run absolutely everything there. The annoying thing about that is that you have to like kind of choose. You got to like, am I doing this locally? Am I in my local cursor? Or am I doing this on the cloud? And there's no real way to like kind of transition from one to another if you're kind of halfway through.

• 34:14Wes Bos?

  So one thing this iOS app now does is you can do like remote control. We've seen this with Claude as well, where if you're like working on something and you have to to leave. And you are running it on your laptop, like a lot of people still are, then you can still open it up on your phone and say, oh, like approve things and whatever when you're when you're doing. I have this all the time where if you're running something and then it's like five o'clock, you got to go make dinner or whatever, but you still want to check a few more things and make sure that doesn't get stuck on it. It's really nice to have.

• 34:49Scott Tolinski?

  Yeah. Why stop working when you could keep working?

• 34:52Wes Bos?

  Exactly. Just work all the time.

• 34:54Scott Tolinski?

  Yeah. Who cares about everything else?

• 34:57Wes Bos?

  Looks like, which is it's crazy to see how many things in the the age of AI, how many things are only iOS these days?

• 35:06Scott Tolinski?

  Yeah, I I use Terminus like somebody else. Judge Mavi says Herder plus Tailscale and Terminus. That's exactly what I use. Terminus is just a terminal app for the iPhone. And again, you just, I can even Mosh. They even have Mosh with Terminus and I can connect via Mosh.

• 35:24Wes Bos?

  What are these words you're saying? What is Mosh?

• 35:26Scott Tolinski?

  Mosh is an SSH alternative that reestablishes connection. So that way, if like, you know, if you have an SSH, you close your computer, you open it, you have to re SSH in. Mosh, you just, it just picks right back up. You don't have to re SSH in. It's just there. And other than that, it's fully the same. So you just replace the word SSH with the word Mosh and it just functions the exact same. So.

• 35:53Scott Tolinski?

  It's sick. Yes. And I so I do that and then I pop open Herder, which Herder is a bit of a challenge on the Terminus, the iPhone app that is a terminal app. You've got to get your keyboard shortcuts dialed in and stuff like that. But yeah, I've been doing that myself for when I need a prompt on the go. And since all of my domains and dev projects are hosted on a little computer back here and served up via my Tailscale network via.

• 36:21Scott Tolinski?

  Caddy to do custom domains for all that that ish. I can just pop open the domain on my phone and see it take place and actually work on it 100% on my phone. If I need to.

• 36:33Speaker 00

  Nice.

• 36:33Scott Tolinski?

  Pretty hot.

• 36:33Speaker 00

  I think we all need to do a setup breakdown because I'm I have a similar setup, but it's not terminal based. It's all like web based tools that do a similar thing. But it's basically like all of these services like cursor. And I think even Claude launched like the way of setting up a remote session so you can continue it on your phone. I think that's where some of these companies are trying to potentially make their money is like make that whole setup and workflow really easy for devs. So they don't have to set up tail scale and like SSH and web servers. Not that it's hard, like it's once you get it going, it's it's kind of pretty easy. But it is interesting to see the companies basically providing those things as a service, that whole setup.

• 37:15Scott Tolinski?

  I don't know what the warp dev terminus is. That might be something different whoever posted that.

• 37:22Scott Tolinski?

  It is.

• 37:23Speaker 00

  Termius.

• 37:24Scott Tolinski?

  Termius terminus. Oh, boy, this is a classic case of dyslexia right here. Sorry about that, folks. Termius, yes, termius, termius, yes.

• 37:35Wes Bos?

  Nice. What is what's termius from?

• 37:38Scott Tolinski?

  It's an iOS app that allows you to connect over SSH or whatever to any term. And you could just use it like a terminal directly on your phone.

• 37:48Wes Bos?

  Okay, okay.

• 37:49Scott Tolinski?

  Oh, you can't see that.

• 37:51Speaker 00

  I'll say if you're on Android, you can use termux. That's what I use.

• 37:54Scott Tolinski?

  Yes.

• 37:54Speaker 00

  Yeah, but it's basically a native app that gives you SSH access and you can have multiple tabs with multiple SSH sessions.

• 38:02Scott Tolinski?

  It's yeah, it's nice. Yeah.

• 38:04Wes Bos?

  Cool, but like the one thing about all of these like terminal based ones is that there's no like, what if you need a browser when you're on?

• 38:14Wes Bos?

  On the go.

• 38:14Scott Tolinski?

  I use my normal browser.

• 38:16Wes Bos?

  On your, on your phone?

• 38:17Scott Tolinski?

  Yeah.

• 38:18Wes Bos?

  Okay, okay.

• 38:20Scott Tolinski?

  I just serve all my stuff via Caddy, so that way I got.

• 38:22Wes Bos?

  Yeah, but like.

• 38:23Scott Tolinski?

  Yeah.

• 38:25Wes Bos?

  But like, what if it needs to like, like puppeteer that, I guess, like, is the kind of the cool thing about a lot of these remote ones is they, yeah, they give you like a they give you somewhere to run your code, right? Like a Linux VM, but then they also give you a browser.

• 38:42Scott Tolinski?

  Well, I got a whole computer that's got agent browser.

• 38:46Wes Bos?

  Yeah, but like, how do you, how do you view that though?

• 38:47Scott Tolinski?

  Chrome DevTools, MCP. How do I?

• 38:51Wes Bos?

  How do you view that on your phone?

• 38:53Scott Tolinski?

  Like, I just have it edit the dev server and then I go back to the dev server.

• 38:58Wes Bos?

  Yeah.

• 38:58Scott Tolinski?

  Yeah.

• 38:59Wes Bos?

  Do you need to?

• 39:00Speaker 00

  Well, to clarify, Wes, so what you're saying is like, let's say an agent spins up a browser to do something. You're asking, how do I see what the agent is seeing? Is that what you're asking?

• 39:08Wes Bos?

  Yeah, yeah, like that's, that's, in most cases, the agent will just spin up a headless browser and it'll do what it needs to do. And you don't care about what that actually looks like. But in, in, in some cases where you actually want to, to see the browser, like, or you want it to spin up a browser and you want to like, log in. That's, that's one thing I do a lot as well. This morning I was making some Yoto cards. And I have a whole bunch of files on my computer and you need to upload them to, to Yoto as a playlist, right? So I just say, hey, spin up an agent browser, let me log in. And then go ahead and, and do your thing.

• 39:47Wes Bos?

  And in that case, it was, yeah, the agent needs to use it most of the time, but I still need to, to be able to dip into it every now and then. And like, I don't, I don't know what that workflow would look like on.

• 39:58Scott Tolinski?

  Here's what you do, Wes. You, you paste in your API keys, your usernames and passwords, your social security number into the chat and you say, here, use this. Yes.

• 40:09Speaker 00

  Yeah.

• 40:10Wes Bos?

  I think that's still an interesting space that that people will have.

• 40:15Speaker 00

  I haven't been to that use case specifically, but personally, if I were doing it, I would set up like a V and C server on my

• 40:21Wes Bos?

  Yeah.

• 40:21Speaker 00

  server. Like that way I can remotely see the desktop instead of just SSH. So that's doable as well.

• 40:27Wes Bos?

  Yeah, yeah, I agree.

• 40:29Speaker 00

  Yeah.

• 40:30Wes Bos?

  Um, can we talk about this Coinbase?

• 40:35Speaker 00

  Do it.

• 40:35Wes Bos?

  Kind of interesting. So we had the Coinbase folks on the podcast in October, which was a couple of months before a lot of this, like, oh, this is actually really good at writing.

• 40:49Scott Tolinski?

  Yeah, it's before the OPUS model dropped.

• 40:51Wes Bos?

  January, like late December, OpenClaw started popping up. Everybody played with it over Christmas. And then once January hit, everybody realized, oh, wow, these models are actually really good. And we were like, I was like, just kind of punching the air because we had the folks from Coinbase on the podcast. And that podcast did horrible. And people were so mad at us about it because at the time Coinbase was saying, we write 50% of our code with AI and nobody could believe it. And it was, it was unbelievable and everybody was mad. And we got probably most thumbs down episode ever.

• 41:28Scott Tolinski?

  People were grumpy about that. You're right.

• 41:29Wes Bos?

  Yeah, they were very, very grumpy about that. And then it turns out like a lot of the Coinbase stuff was, was right. Now Coinbase has now come out, like everybody over the last was like more tokens, more tokens. Who can use the most tokens? And then you realize, oh, this is really expensive. Maybe the, don't spend as many tokens as you possibly want. So Coinbase, obviously spending a ton of money on tokens has said, like, I don't, here's your unlabeled graph of the day for everybody to look at. Obviously, the amount of tokens that they had been using has gotten. Like exponential. And they said that they have been working on better defaults.

• 42:11Wes Bos?

  So they can route a lot of their stuff through. We're experimenting with defaulting open weight models like GLM 5.2 is what CJ was just talking about and KIMI 2.7 through our LLM gateway. While we still encourage engineers to choose the right model for the text, 91% of our employees were never hitting the usage cap. So instead of lowering caps and driving up alerts, we're moving to cheaper defaults. Note that code reviews use a diversity of models so they can check each other's works. So basically they're working on better routing, being able to, I don't know, I think they're doing smarter routing here, being like, if you ask it something like, hey, can you LS the files in this directory?

• 42:53Wes Bos?

  You don't need to be sending that over and spending $9 on a, on a single prompt. So it looks like they're investing in their own infrastructure.

• 43:03Scott Tolinski?

  Me asking.

• 43:04Wes Bos?

  Gateway.

• 43:05Scott Tolinski?

  Opus or me asking GPT 5.6 Luna Ultra to LS some files.

• 43:13Wes Bos?

  Putting this into practice has cut our AI spend nearly in half while our token usage continues to grow. This is kind of super interesting to me is like, yeah, these now companies are spending time on their infrastructure. Where they're like, yeah, some stuff needs to go to Opus. Some of the stuff we can be sending it over to some of the open weight models and we can run that ourselves. So love to see it. Yeah.

• 43:39Speaker 00

  Yeah, I, I think we're going to see more companies doing this and potentially companies spinning up that do this for you. Like basically wrappers on top of open router, but it's like a, no, it's like almost zero config. You can just start using it at your company and your, your bills are lower because it's using the right models or choosing better models.

• 44:02Speaker 00

  I did, I did find something just yesterday. I haven't had a chance to try it yet. But it is called Wayfinder Router. And it basically allows you to do this yourself locally.

• 44:16Speaker 00

  Waiting for my screen share to pop up. Am I sharing my screen? I think I am. Yeah, there it is.

• 44:20Scott Tolinski?

  Yes, there it is.

• 44:21Speaker 00

  So this is on GitHub. It only has 250 stars. It's pretty new. Again, I just found it yesterday. There's probably other tools like this, but when I read this readme, I was like, this is it. So it says it's a fast, offline, hard or easy call on every prompt, score deterministically with no model call. Route the easy ones to your small or local models and the hard ones to your big one, or compose any model router behind it. And the idea is you would use it almost like you would Claude, but basically it's like a proxy. Every request goes through its filter first to figure out where should it send the response, shouldn't send the request? And then it goes off. But what I found interesting is that they've come up with deterministic rules.

• 45:00Speaker 00

  So it, instead of, a lot of, a way that a lot of these routers work, it'll talk to a really cheap dumb model first that says, given this prompt, which of these models should I send it to? But this looks at the complexity of the code. So I'm see if I can find the section on how it works. But it's essentially like, if the prompt is of a certain length, go ahead and send it to a bigger model. If a prompt includes code, send it to a specific model. You can basically have deterministic rules that choose the model. And to me, that's a lot better than sending it to an AI, which sends it to an AI.

• 45:37Scott Tolinski?

  Yeah.

• 45:37Speaker 00

  And I'm excited to try this out with my local models because I can basically have like a really small model for simple things and then also run a 70B model anytime I need more expensive things and that's all handled fairly. So.

• 45:52Scott Tolinski?

  Yeah. And the best, too, you could need fast models for fast tasks. Like sometimes, yeah, you ask it to do something that you'd hope would be a fast task and it's like thinking about it, you know?

• 46:03Wes Bos?

  Yeah, it's going forever. I'm so curious to see if these, like, like if this actually works well. Being able to determine if the output of it is, is good or not. You almost need like a, its own model to determine where to, where to send things, you know? Because, yeah, how do you know? If like, if you give it a huge code prompt, it might be something that a cheap model can fix in a heartbeat, but it might not, it might, might need to catch something. So it's really tricky, but I'm, I kind of think I want to try this as well.

• 46:43Speaker 00

  Yeah, I want to try it, just, just to see how good it is. And I was reading their CLI also has like a dry run. You can basically paste in a prompt and see what complexity score it gives it.

• 46:51Wes Bos?

  Okay.

• 46:52Speaker 00

  Where it would have routed it.

• 46:53Wes Bos?

  I love a dry, I love a dry run. Um, I, I do, I, I like, there was a, like, Fallow is one that I rely on a dry run a lot for, or those types of, any type of auto fixing tool.

• 47:06Speaker 00

  Yeah. And I guess for anyone that wants to look into this space, they have a comparison table. I haven't used any of these, but, but I want to try at least Wayfinder for sure. But there's Route L L M, um, there's Open Router, there's Light L L M. All of these are a similar idea where it basically routes your requests to specific models. So.

• 47:31Speaker 00

  Yeah. And in the chat, Rave is saying that Pi has a router extension like this. So if you're using like the Pi, um, harness, that potentially has a plugin you can use to, uh, choose the right model as well.

• 47:46Wes Bos?

  Awesome. Uh, we lost Scott, so this stream just got way better.

• 47:53Speaker 00

  Just me and you.

• 47:54Wes Bos?

  Well, should we talk about when Scott's gone here?

• 47:56Speaker 00

  Uh, let's talk about, um, Tailwind. Yeah. Any news in the, in the world?

• 48:03Wes Bos?

  No, I haven't, haven't heard much. Actually, here, I've got something here. Um,

• 48:09Wes Bos?

  let me share my window.

• 48:14Wes Bos?

  Speaking of token efficiency, uh, Martin Woodward, he, I believe he is VP of Del Verell for GitHub.

• 48:25Wes Bos?

  Was saying, everybody argues about models, but we've always known this. It's the harness that really matters, especially when you have full choice over the model. Want the data on how GitHub Copilot compares and how it gives you better results using less tokens, token efficiency. This is wild to see that with all of the same, like Opus 4.7, Sonnet 4.6, GPT-5.4 and 5.5, which one uses less tokens? Either the native model, like Codex or Claude, or the Copilot CLI? And they're saying, in many of these cases, Copilot uses like up to, what, 50, 60% less

• 49:08Wes Bos?

  tokens than the native harness? And whether or not that gives you, I'm curious if that gives you as good of output. But you have to think that GitHub Copilot cares a ton about token efficiency because they're paying for the API costs here, whereas like Claude Code, maybe will use a little bit more tokens because, you know, it's their own, it's their own supply. But if they, if GitHub Copilot can cut the token usage in half and still get similar outputs, then their, their bill obviously is not going to be nearly as large. And that's got to be a huge expense for GitHub right now.

• 49:49Speaker 00

  Definitely. It's fascinating to see. I mean, you almost wonder, like, if you are using the harness that comes from the provider itself, like, are they, I guess, I can't, I haven't looked at this in detail, but are they using less tokens or more tokens? And is that?

• 50:07Wes Bos?

  Apparently way more. That's what this whole chart is.

• 50:10Scott Tolinski?

  Part of that too, is like, so, like the default prompts and stuff. So back a few weeks ago, I was annoyed that Codex kept adding gradients to things. And I found that the default prompt in OpenCode was to say to use gradients.

• 50:29Wes Bos?

  But DAX said specifically that that came as a direct copy and paste from the Codex Harness. So they wanted to keep it as one to one as the Codex Harness. So Codex is telling it to use gradients and stuff. It's all additional system prompt stuff.

• 50:48Speaker 00

  Yeah. So I would say like, put your tinfoil hats on because they're doing this, they're potentially doing this because you eat into your usage faster and then you don't get as much as you could for the price that you're paying. Possibly, I mean.

• 51:04Wes Bos?

  You think so?

• 51:05Speaker 00

  I don't know. I mean, there's, I don't, it could be that, but I would say, don't attribute malice to stupidity because it could just be that the prompts are bad, the harness is bad. The harness prompts have way too many tokens in them, way too, way too much context that.

• 51:20Wes Bos?

  Yeah. Or that they, they want a more verbose trail because it's better for training, you know? Maybe they, they care, they don't necessarily care about the token usage. I don't think any of these companies are at a spot right now where they're like, let's get people to use more tokens, you know? They're, everybody's strapped for compute. I think a lot of them are just like, let's get as many paying customers in the door right now and we'll worry about the profitability in the, in the coming months as we IPO.

• 51:50Speaker 00

  Yeah, definitely.

• 51:53Wes Bos?

  Hmm. Yeah. I don't know. I tend to follow less on the cynical side of things in that regard, but I'm also dumb. So you know what? It could just be that.

• 52:07Speaker 00

  Nice. Well, I've got something else to share if we, if we want to keep moving.

• 52:10Wes Bos?

  Yeah.

• 52:11Speaker 00

  Let's do it. And if you're just joining us, Syntax Weekly, we're talking about tech news, web dev stuff, all the things happening in this wonderful world of ours. And we're getting close to the point where we're going to be taking a look at things that people share in the chat. So if you have any interesting news or libraries, libraries, or whatever else that you want us to talk about, be sure to throw those in the chat. But this is a UI library I came across called Performative UI. And the idea is, it's a bit of a joke, but at the same time, the components are really cool. But they call themselves AI-native React components for basically AI startups. And so it's basically tropes that you see in a lot of AI startup websites.

• 52:55Scott Tolinski?

  Very fun.

• 52:55Speaker 00

  So first of all, they've got ASCII hero. It's got like this really cool like ASCII visualization and you can hover over it and it like highlights where your mouse is. But you can just now drop this in as a React component. I mean, as tongue in cheek as this is, it's also really cool. Like this is a cool little visual to like throw on your site.

• 53:17Scott Tolinski?

  Yeah.

• 53:17Speaker 00

  Another one they have, I love this one, the Golden Eye Effect, where like when you hover your mouse, it creates a circle and reveals something else behind it.

• 53:26Speaker 00

  So that's like, I mean, not crazy hard to build, but the fact that you just have a component you can drop in is pretty cool.

• 53:34Speaker 00

  And then they also have, what is it, Word Roll? Yeah, so you know how it's a sentence where the last word kind of like flips into a different word?

• 53:46Speaker 00

  They have a component for that as well.

• 53:48Scott Tolinski?

  Oh, this is very funny.

• 53:49Speaker 00

  Yeah, this is a fun little find. I found it like last week on Hacker News. There's tons of other components to dip into. I almost want to give AI this site and be like, let's launch an AI startup, build the site with this and go crazy and see all the things that it adds in there.

• 54:08Scott Tolinski?

  My God, my God, this is sick.

• 54:12Speaker 00

  It's fun.

• 54:13Scott Tolinski?

  Can I show my robot?

• 54:15Speaker 00

  Yeah, let's see it.

• 54:19Scott Tolinski?

  Oh, his little hat, whoa, what happened to your head there, little guy? Uh-oh.

• 54:22Speaker 00

  Uh-oh.

• 54:23Scott Tolinski?

  Uh-oh.

• 54:24Speaker 00

  He's dead.

• 54:25Scott Tolinski?

  Oh, no. This is, we call him Marshmallow.

• 54:31Wes Bos?

  Let's talk about Sentry real quick while Scott fixes that. Cause a code breaks, Sentry can fix it. And I guarantee you take the logs or whatever is going on with Scott's dead little robot right there. I bet Sentry could get that thing fixed up in no problem. It'd be able to, you throw a seer on there, get down to the root cause of what actually happened. Send a pull request for actually fixing it. Sentry is excellent. If you're building any, any type of application, you know, backend, frontend, full stack, iOS, whatever. You need to make sure you have Sentry in there for monitoring, replay, traces, logs, absolutely everything for fixing your application.

• 55:16Wes Bos?

  Go to sentry.io forward slash syntax. That's going to get you two months for free.

• 55:23Scott Tolinski?

  We can tell by the dashboard that I've built that there's something currently not quite right with the software side of things. Needs attention. Robot unresponsive. It's not the robot, sadly, it's my software.

• 55:38Wes Bos?

  Yeah, what is this robot?

• 55:40Scott Tolinski?

  Okay. The robot is the Reachy Mini. It's from Hugging Face and Pollen. And it is really super cool. It, it basically onboard, at least this version of it. It allows you, whoa, he's doing something. Basically it allows you to have what could be like a Google Home or something that has a camera into it. Oh, he's cooing, but his head's not up. Why isn't your head up? Either way, he's got, he's got a camera, he's got speakers, he's got a base, he can move around. You could tell him to look at stuff, he could talk to you. I'm going to be using this for a lot of fun projects with my kids since this comes in like a thousand different parts and like a 60 page instruction manual.

• 56:23Scott Tolinski?

  So we, we've been working on this together. We're building like a spelling app for it right now. But I have been writing all of the software myself and by that I mean AI has been writing it for me. And because of that, it is currently breaking all of the time. Because, yes, I am Weekend at Bernie's, this robot, I love that movie by the way. But it's sick. It's so much fun. The kids have been having a blast with it and he's very expressive typically, although he's, he's playing a joke for y'all right now. He's being all. All uh shy, but yeah, it it's pretty dope. I'm I'm do a dedicated video on it right now. I I show us putting it together.

• 57:02Scott Tolinski?

  I'm talking about the software we're building. I'm building an app to have it like help our kids with spelling tests instead of a camera. You can point it at their spelling tests. And then they can practice spelling back and forth with it. Uh, reason being is that our kids get really frustrated during spelling tests and spelling practice. But I would imagine if they had a cute little robot being like, oh, that's not quite right. Uh, they might have a lot more fun with it than me being like, hey, can we try again? You know? Uh, so there's a lot here. I'm gonna also have a um, use the camera to help us learn how to play Warhammer 40k as well.

• 57:36Speaker 00

  Yeah.

• 57:36Scott Tolinski?

  Uh, there's a number of projects with it. But yeah, like I said, I'm writing my own software for it instead of using theirs. So, uh, currently it's iffy right now because of that.

• 57:47Speaker 00

  Nice. So what is the, I don't know if you talked about it. What's the the language? Is it running a Raspberry Pi? Can you use anything? Or is it embedded? Do you have you see?

• 57:56Scott Tolinski?

  You sleeping?

• 57:58Wes Bos?

  I I think Scott's, oh.

• 58:02Scott Tolinski?

  Yeah, he's sleeping.

• 58:03Wes Bos?

  Didn't you say like there's two versions? There's one that like you can put a Raspberry Pi in it and it and then there's another version which you have, which you can just like relay the commands uh over wifi from your computer, right?

• 58:14Scott Tolinski?

  Yeah, there's a wifi module and so all the compute happens in a demon that's running on my uh, my Mac mini over there that connects to it via this app. They have their own app uh that has a bunch of UI that you can or or apps that you can just quickly and easily install and everything like that. So um, if you're not like the type of person who wants to hack through something like that, it's still very much a hobby project considering again, the instruction booklet uh is significant. And if you're not comfortable with uh boards and a, you know, 60 page instructions on putting something together like this, then it's it's definitely not something you just get

• 58:53Scott Tolinski?

  and have it work out of the out of the box. So um, it's it's pretty sick though. So I I'm excited to see, give me another couple of weeks here. I'm gonna have something crazy going on with this bad boy.

• 59:05Wes Bos?

  Man, we, we went out for dinner at a sushi restaurant a couple of nights ago and they have these robots that like deliver you the sushi. You order it on the app and like the robot brings it to your table. Have you guys seen that? Any, any of your local restaurants?

• 59:19Scott Tolinski?

  No.

• 59:19Wes Bos?

  No? Oh, it's pretty common here. I, I've been to like probably six or seven restaurants locally that have that. And it's my son, it, it was my son's birthday. So they put like a cake on this thing and then it started singing happy birthday and it like drove its way over the table and my younger son lost his mind. He was terrified at the happy birthday robot. It was so friendly.

• 59:47Scott Tolinski?

  Yeah. I, I, oh my God. He needs to stop making cute noises over here.

• 59:55Scott Tolinski?

  I'm, I'm just like really, really interested in, in, in general, the types of things I can do with the kids in this thing. And like I said, um, oh, sorry. My, my custom software that I'm doing for this, it's all, the reason, one of the reasons why it's so iffy is because it's all based on Hermes. So rather than their version is all connecting directly to OpenAI or whoever, um, mine is like routing everything through Hermes. So that way it has a memory and it has all these tools that Hermes has access to or skills or whatever. So, um, it's a bit more complex than a, uh, just a basic chat back and forth. So again, I can't wait to show off some of what I'm doing here, but of course it stops working the moment that I gotta show it to you guys. So.

• 1:00:41Speaker 00

  As always.

• 1:00:42Wes Bos?

  He's shy. It's okay.

• 1:00:43Speaker 00

  Yeah, he's shy. Nice. I do want to mention, uh, AOA raided us with 14 people. So shout out to AOA. Thanks for bringing all your folks over. If you are just joining us, Syntax Weekly, we're, we're talking about the news. Uh, we're nearing the end though. I only have like one more thing to share and I think we'll get into anything that the chat would like to share. So throw some links in the chat, um, and we'll get into that.

• 1:01:07Scott Tolinski?

  Well, before we do that, folks, how do you pronounce demon? Damon? Demon? We've, we've gone through this before, but how do you pronounce it? I want to know in the chat.

• 1:01:15Wes Bos?

  Damon my whole life. Demon sounds a bit too scary for me to have that running on my computer.

• 1:01:21Speaker 00

  Yeah, I say demon, but every time I see the word, I want to say Damon.

• 1:01:24Scott Tolinski?

  Thank you, Court. I got a salad delivery.

• 1:01:28Wes Bos?

  Oh, man.

• 1:01:28Speaker 00

  Nice.

• 1:01:29Scott Tolinski?

  This has got some rotisserie chicken.

• 1:01:30Wes Bos?

  Salad report. Holy smokes.

• 1:01:33Scott Tolinski?

  Yes, yes, yes.

• 1:01:36Wes Bos?

  Now I'm looking at food delivery robots on eBay. This is not good.

• 1:01:41Speaker 00

  We've got three votes in the chat for demon and one for demo. I don't know how you would say it in French, but.

• 1:01:46Wes Bos?

  Demo.

• 1:01:50Wes Bos?

  Demo. I don't know. I start the demo and it delivers the thing for me. It shuts down, the demo starts again.

• 1:02:02Speaker 00

  Nice. Yeah.

• 1:02:04Wes Bos?

  I'm allowed to make fun of French people, by the way, because I'm Canadian.

• 1:02:08Scott Tolinski?

  I'm also largely French Canadian, in case people don't know this.

• 1:02:15Wes Bos?

  Let us, let her rip.

• 1:02:17Speaker 00

  Sweet. So last thing I have to share is just a fun release for Android. So I'm not in the world of iOS. On iOS, does their built-in keyboard support swiping?

• 1:02:28Wes Bos?

  Yeah.

• 1:02:28Speaker 00

  Like, yeah. And like, that concept goes way back. There is the swipe keyboard that was, this company was founded in 2002.

• 1:02:35Wes Bos?

  That's a good idea.

• 1:02:37Scott Tolinski?

  Yeah.

• 1:02:37Speaker 00

  Yeah. Like, the whole idea is you can swipe to choose the words you want. And at this point, it's commonplace. Like, iOS has it. Android has Gboard. So this is the official keyboard from Google. And it has swiping built in. But if you are using the Google keyboard on your Android phone, it's basically just a keylogger, right? You could, like, the data, the data that Google collects is technically encrypted. Encrypted, it's associated with your account. But literally, every single person using the Gboard keyboard is keylogging every single thing they type, no matter what app it is. And I guess they have like an incognito mode, but like, everything is sent to Google servers.

• 1:03:19Speaker 00

  And that makes it decent, because it provides really good suggestions about like commonly used words and stuff like that. But as someone that's a bit more privacy conscious. I've been using a keyboard called FUTO, F U T O. And for the longest time, their swipe input sucked so bad because it's a fully open source keyboard, built from scratch, and I would try to use the swipe. It would be really annoying, and so I kind of just defaulted back to manual input. But the nice thing about FUTO is they also support local voice models. So if you like to use voice to text, that can happen all on device and download the models. But they recently, with their latest 0.1.29, they released a new swipe-based local model that's really good.

• 1:04:03Speaker 00

  I've been using it, like, I didn't even, I recently got a Pixel 10. I just switched from the Pixel 9 and didn't realize that this was the newer version. And I was like, why is swipe so much better? But they launched this, and basically, the way that they did it was collecting data from users. So I think it started in March of 2025. They released a dataset of a million swipes. So basically, people could go to a website and swipe, and then it would collect the data of like, did it get it right? What motion did they use? And now they also released that dataset so other people can use it to build their own models.

• 1:04:40Scott Tolinski?

  Ooh.

• 1:04:41Speaker 00

  But this is exciting for me because finally, I have a decent swipe that isn't keylogging me and gets things right, and it's fully local. So if you're on Android, definitely check out the FUTO swipe keyboard.

• 1:04:54Scott Tolinski?

  And if you're on iOS, you're doomed to never have a good keyboard ever because Apple doesn't care about the keyboard, yeah. And they won't let you use real third-party keyboards. That's the worst part.

• 1:05:04Wes Bos?

  All the third-party keyboards suck.

• 1:05:06Scott Tolinski?

  Well, that's because they're just reskins and blah, blah, blah. They're terrible. And that was the cool thing, Android, the third-party keyboards were actually worth using. So yeah, shout out to good third-party keyboards.

• 1:05:19Speaker 00

  Yeah. Alright, let's get into some chat things.

• 1:05:24Wes Bos?

  Yeah, I got one here.

• 1:05:25Speaker 00

  Yeah.

• 1:05:26Wes Bos?

  Sorry, go ahead.

• 1:05:26Speaker 00

  But links are blocked on YouTube, so if you want to share something on YouTube, just type it out. We'll find, we'll find the link if you just type it out. Or you could head over to Twitch, you can share links on Twitch, but yeah. What do you got, Wes?

• 1:05:37Wes Bos?

  This is CSS Quake. So a couple streams ago, somebody sent us, what was it? CSS Poly, which is the ability to do 3D rendering.

• 1:05:48Scott Tolinski?

  Yeah, and there was CSS Doom already.

• 1:05:50Wes Bos?

  So this is CSS Quake, which may not be impressive to you, and I can't open my dev tools while it's playing, but all of the textures and the whole space of Quake is being rendered out. It's actually glitching out here. Is being rendered out fully in HTML and CSS, meaning that like every time you see something that is 3D, it's literally just like a transform, translate, z, whatever in CSS. Like all of these, everything you see in here is an HTML element with CSS applied to it. This is no canvas, no 3D, any of that, which again, we don't know why anybody would ever want to do this, but the fact that someone figured it out, unbelievable.

• 1:06:37Scott Tolinski?

  Man, do you guys play Quake? Quake was a dope game.

• 1:06:41Wes Bos?

  I know.

• 1:06:42Speaker 00

  I did.

• 1:06:42Wes Bos?

  I, I didn't really play much.

• 1:06:44Scott Tolinski?

  You guys are crazy. I cannot believe we're all the same age-ish and you guys didn't play Quake. That's wild.

• 1:06:50Speaker 00

  Not much of a gamer, I don't know.

• 1:06:52Scott Tolinski?

  Oh my god.

• 1:06:53Wes Bos?

  I just got into a basement, look at me. I'm so good at this.

• 1:06:58Scott Tolinski?

  Oh my god.

• 1:06:59Speaker 00

  You should be a gamer, Wes.

• 1:07:00Wes Bos?

  Dude, maybe I, maybe I'll, I'll change. Screw web development and screw software.

• 1:07:06Scott Tolinski?

  Game streamers? I could do game streaming, yeah, I could do that.

• 1:07:13Speaker 00

  Yeah.

• 1:07:14Scott Tolinski?

  I don't think you'd make those kind of noises in Quake. I think typically Quake is more of like a you know, pew pew pew type of, yeah. It's a little bit more, uh, robust and manly than that was.

• 1:07:26Wes Bos?

  What else do people send in?

• 1:07:28Speaker 00

  I've got one. They mentioned this earlier in the chat, so I'm letting them go first because they reminded us. But they wanted to know our opinions on the Invoker Commands API. So do you guys have any experience with it? Can you tell us what it is? Is there any new news around it? But yeah, Invoker Commands API, what is that?

• 1:07:44Wes Bos?

  Yes. Go ahead, Scott.

• 1:07:47Scott Tolinski?

  Okay, Invoker Commands, basically being able to fire off events that would typically be reserved for JavaScript with HTML. And out of the box, there's just like a handful of ones. So like if you want to toggle a pop over, those types of things, that's what's going to be in the Invoker API right now, today. And it's a, it's a dope API. And it's available, let me check, I don't want to say anything dumb here, but like-

• 1:08:17Wes Bos?

  Baseline 2025.

• 1:08:19Scott Tolinski?

  Yeah, baseline 2025. And, and so for the ones that are actually in the browser right now, I believe it is just pop over. I used to have these memorized.

• 1:08:30Scott Tolinski?

  Gosh, MDN.

• 1:08:31Wes Bos?

  Pop over for dialogue.

• 1:08:35Wes Bos?

  And what's the other one?

• 1:08:37Scott Tolinski?

  Gosh.

• 1:08:37Speaker 00

  I had no idea that got added. So that basically means you can open and close a dialogue without JavaScript, you know?

• 1:08:43Wes Bos?

  Correct. Exactly. That's the whole point of this API is that you can declaratively make interfaces. So you can have a button and another element. Another example that doesn't exist yet, but you can, you can make your own custom ones is that if you have a video player and you want to have a button that pause, plays, skips ahead, any of these types of things, a declarative way that you could simply just wire up this interface is, is like, like you normally would like make a click handler, right?

• 1:09:15Scott Tolinski?

  Yeah.

• 1:09:16Wes Bos?

  And then now you're in HTML land and-

• 1:09:20Wes Bos?

  JavaScript land.

• 1:09:21Scott Tolinski?

  Yeah.

• 1:09:21Wes Bos?

  But it's so nice to be able to just register your own invoker APIs and then you're fully declarative.

• 1:09:28Scott Tolinski?

  So the ones that are in the are shipped right now are show hyphen modal, close, request, close, show, pop over, hide, pop over, and toggle, pop over. You can also create custom ones as well. However, like custom ones, I found that people are like, I wonder why you would even do that. If you could just, you know, do listeners and stuff. But the ones that are coming, like Wes said, would in fact be things like media elements, which, man, I would think that that would rip. There's a whole bunch of potential future ones that I tried, there used to be a nice organized list of this stuff on

• 1:10:07Wes Bos?

  Yeah.

• 1:10:07Scott Tolinski?

  But they've since

• 1:10:08Wes Bos?

  Even like opening up like a customizable select. You know, if you click on something that is totally somewhere else in the in the browser and you want to be able to open that up. It's so nice to be able to just like, like, they don't even have to be built into the browser. I have no problems registering my own commands. But then having like a standardized way, because I've, I've written this code for many years with data attributes, you know? Data dash command is equal to rotate left or rotate right or pause or toggle. Media, volume, all of those types of things is so nice to be able to just simply wire them up with the standardized invoker ones, especially when like,

• 1:10:55Wes Bos?

  you also may have like third party libraries that are exposing the functionality. And for you, the user, it's no longer, okay, import this JavaScript, add an event listener or whatever. Simply just put a command and a command for attribute on your element and you're, you're up and running.

• 1:11:14Scott Tolinski?

  Yes. Proposed ones are enter or toggle, full screen, copy text, share, play, pause, show picker for select lists, close, open and close or toggle for the details and summary. Yeah. If we could get a copy text and share as invoker commands, hell yeah.

• 1:11:35Wes Bos?

  Copy text, yes.

• 1:11:37Speaker 00

  No JavaScript for like clipboard, add to clipboard, right? That'd be crazy.

• 1:11:40Scott Tolinski?

  And even, I mean, people like, folks, you shouldn't be using a library to do that today. You could, like, the API is so easy in the browser, but people still do that anyway. So, yeah, but it would be dope to have that as just a quick little command.

• 1:11:54Speaker 00

  A good example, Stoney mentions is it throws away all the library work to make a nice customizable customizable video and music player in the browser.

• 1:12:01Scott Tolinski?

  Yeah.

• 1:12:02Speaker 00

  So every single button, like play, pause, fast forward, change the speed, all of those, right now, you have to use JavaScript to like change the video object, but eventually each of those could be just a command that it invokes and then you need no JavaScript for a custom controls UI for for a custom video player.

• 1:12:20Scott Tolinski?

  Yes.

• 1:12:21Wes Bos?

  That's the, um, what's the one that I use? Muk, what's the one that Mux makes? Why am I, I'm blanking on?

• 1:12:26Scott Tolinski?

  Media Chrome.

• 1:12:27Wes Bos?

  Media, Media Chrome. That's what Media Chrome is, you know? They have elements for everything and you want to make your own custom media player? You simply just slap the elements that you want in there and they all just look up to the parent scope and are able to work from there. You can of course write your own JavaScript, but in almost every single use case, there's an element for what you need and you can make your own by simply just writing HTML, right?

• 1:12:55Scott Tolinski?

  Yeah.

• 1:12:55Wes Bos?

  Or writing custom components.

• 1:12:57Scott Tolinski?

  Yeah.

• 1:12:59Scott Tolinski?

  Word. Cool.

• 1:13:00Speaker 00

  Sick. I got one more from the chat if we want to talk about it.

• 1:13:04Wes Bos?

  Yes.

• 1:13:05Speaker 00

  And because I had it on my list, I just didn't want to bring it up because we always talk about CloudFlare, but Alex shared it in the chat, so it's from the chat. But last week, they released, CloudFlare released temporary CloudFlare accounts for AI agents. And the idea is if you have an AI agent. Working on some code, building a site. It can actually deploy to Cloudflare without an account. It uses the -- temporary flag on Wrangler, the CLI. And that makes a deployment live for 60 minutes, in which time you can claim that temporary account. And this is, it's, first of all, it's interesting. I wonder how long this is gonna last before too many people create so many temporary accounts that they're like flooding Cloudflare servers.

• 1:13:49Speaker 00

  But if you think about the use cases, I feel like people that maybe aren't familiar with how to deploy, if they can get this into the training data and people become aware of using this temporary deployments for it. Now, an AI agent might direct like a more non-technical user to be like, "Oh, let's just use Cloudflare to deploy."

• 1:14:09Wes Bos?

  Yeah.

• 1:14:09Speaker 00

  I definitely, I see this as a way of getting non-technical people into the Cloudflare ecosystem, as well as a way for just your random one-off agents to be able to try things out and deploy them without having to link it to your account and your billing and everything else like that.

• 1:14:24Wes Bos?

  Yeah, you wanna just put something on the internet temporarily, even if it's just so that you can pipe that into another agent. That's a good spot to put it. This, this just seems like the, you can add stuff to cart before you're signed in and then, and then you can go ahead and sign in later. You know, like, or like, this is just like the, the free trial extended. It's absolutely genius from them because of, of what it means for onboarding, meaning that now you can, like, like, like you said, CJ, it's, once it gets in the training data, this thing can, you could just create something and people go, "Wow, the thing that I had an idea of, I typed into the box and now it is on a live thing."

• 1:15:08Wes Bos?

  And if you want to share that with your friends, then you gotta sign up for an account. I think that is going to be absolutely huge for getting new customers come their way.

• 1:15:20Speaker 00

  Yeah, and I wouldn't, so it says, "Claim account." I wonder if it's gonna let you just claim the project, right? Because what if an agent spun this up, but you already have a Cloudflare account?

• 1:15:30Speaker 00

  I don't see it, but that would be good. That would be good to have.

• 1:15:34Wes Bos?

  Yeah, that's interesting. Yeah, it's an interesting one. It's like, "Yeah, I already made this, but I made this on like, um..."

• 1:15:43Wes Bos?

  Yeah, like, like even now, we're talking about people doing remote development, you know, if you don't have your Cloudflare MCP set up on that whatever box that you're working on and it deploys that thing, I guess it could just redeploy the whole thing once you are signed in. But yeah, being able to transfer that project over, maybe, I don't know, the domain transfer story on Cloudflare is kind of annoying. I don't know if they'll, they'll get to that.

• 1:16:13Wes Bos?

  Cool. I didn't realize that they did this. This is neat.

• 1:16:17Scott Tolinski?

  Yeah, no, sick. I didn't either. Guys, do we have anything else that we want to hit before we get on out of here?

• 1:16:24Wes Bos?

  I got it all. That's good for me.

• 1:16:27Speaker 00

  Yeah, I think we're good.

• 1:16:28Scott Tolinski?

  Sick. Well, thank you, everybody so much for joining us. We will be back on next week. I think Wes will not be here, though, will you, Wes?

• 1:16:36Wes Bos?

  I won't be here. Nope. See you suckers later. I'm gonna be not thinking about agents and just relaxing.

• 1:16:43Scott Tolinski?

  That's right. Although I did hear that you have all your agents being able to access from anywhere. So you're gonna be, once you shut that laptop.

• 1:16:50Wes Bos?

  Well, yeah, I'll shut the laptop and open my phone. So don't worry about that. Yeah.

• 1:16:55Scott Tolinski?

  Sick. All right. Well, thank you so much for joining us. We will catch you in the next one. Peace.

• 1:17:00Wes Bos?

  Peace.