Roundup - Sick AI Tools For Developers

Boss, and Scott

Welcome to syntax. On this Monday, hasty treat, we're gonna be talking about using AI tools for development and just kind of what types of interesting things are out there, what are some interesting tools that you might not have heard of, some of them you may have heard of, Just kind of tips and techniques and interesting things to look at and and really just try out. My name is Scott Talinski. I'm a developer from Denver, and with me as always is Wes.

Wes, what's up, man? Hey.

Not too much.

I am kinda excited to talk about this because There's a whole world beyond GitHub Copilot, and it seems like every week there's some sort of new thing that are popping up. That's the kind of weird thing about being a developer is it's almost always developer tools that surface first in new areas of tech.

So there's a lot happening right here, and it means that we'll probably see it ripple out into many different areas of Both coding as well as just, like, in life, like, what was I doing today with the this is a bit of an aside, but let's do it.

See you. Today, I was trying to find out if a 5 eighths bolt and a 16 millimeter bolt were the same, Meaning that, like, 5 8 inches, 15.6 millimeters.

And I was like, does that 0.4 millimeters actually matter? So I just asked AI and this is, yeah, they'll fit, you know, rather than than going into it. And also like I bought a TV at a yard sale This weekend, and there's there's a sticker on it. It says works sometimes, and I was like, it's calling my name. Sometimes? That's I bet I can make it work all the time. So I started googling around for the issue that it had, and there's so much garbage SEO juiced articles of like, Here's 9 things to test. And then you get all these YouTube videos that are like Generated are people that try like the stupidest things, like turn it off and turn it on. It's like, of course, I've tried that. Of course. Yeah.

But But then I was like, you know what? Screw this. I just typed all the symptoms into the AI, and it came out with the answer that I think is right. We'll see. But I was just like that cut through all this BSE SEO clickbait, 57 ways, eHow, All these different things that was so annoying. And I was like, this is 2 examples of where AI is starting to creep into Things that are not development.

Yeah. Yeah. I know. It it's I totally agree. I mean, I was even, like, getting it Up and running to get an SSL certificate for my Synology. Yeah. I want to get an SSL certificate for Synology, I haven't want to have it work with my media.

I want to have it work with all these things. I want to expose those ports specifically To the greater web, to use them remotely, like, how do I do this? And it gave me, like, a step by step list. Exactly. Do this, then do this, then do this. And yet all of those instructions, it turned out by the end of it, it was all working and secure correctly. Well, let's talk about some things that you can do with these tools in development.

And we're gonna be talking about a handful of tools, extensions for Versus code. Sometimes they're just services. Sometimes they're things that just plug in directly to what you're doing.

You know what? I I have been recently trying out a lot of these tools inside of Versus Code for me. Anytime anything pops up, I'll I'll give it an install.

So my little activity bar on the side is now 8 or 9 different AI driven, like, little chat windows directly in my code. I'll just different extensions Just to see if anybody's doing anything differently. And one of the ones that I I came across was CodeAI.

And Kody AI is an extension by Sourcegraph.

And so the thing that Kody does well is you Upload your repo essentially to source graph.

Then anytime you're asking questions of Cody, it's asking always within the context c s of your code base. So the idea is is that this tool knows and understands the context of everything it's working on. And I found it to be pretty neat. There's a lot of what are they called recipes where you can do things like select some text and say, write me A, type or j s doc style comment for this, and it will just do it.

Or give me a code smell.

So if you select some code, you can say smell this code, and then it lists out some things that it you might want to think about. And they're not necessarily in terms of, like, Fix this, fix this, fix this. But it's like, here are some things that you're doing that are potentially have issues and, you know, whatever.

And I found that to be really, really useful.

Especially, like, asking it, is there a better way to do what I'm doing right now? Yeah. Can you optimize this code? That was, like, a whole another recipe in there was, just for a test, I did like an if statement, like multiple if statements, and I said optimize this code. And it converted it to a switch statement, and then it gave me a paragraph about why this particular switch statement executed faster than the if statements I had In big o notation. Wow. And

so it's both a Versus code plug in, but it also seems to be a,

Like something that you log in to that has a whole UI? Yeah. And I oh, my experience with the logging into of Sourcegraph was just So that the plug in would work. Oh,

you know what? It seems like it's an AI for your business, and one of the parts of your business is coding. Right? That's a good name. I think that might be the best name. What, Cody?

Yeah. Yeah. What do you think? Yeah. Well, the only reason why it's not is Because if you look at CodeAI, there's like 8 other companies that are like, oh, to be CodeAI. And it was like hard hard to find the right one when I initially searched for it. I wonder how this works because,

so for everyone listening, the way that a lot of this AI stuff works is You have to give context for what it is that you're asking. And so GitHub Copilot will take a bunch of context cues when you are when it's trying to both autocomplete it as well as when you're using the new chat, it will say, alright, well, the current tab is named this. The other tabs are named this, and It might upload a bit of code before and after, and it tries to get as much context for what you're doing.

I don't know what else it does. They haven't peeped about that, but I'm assuming it would also maybe look at your ESLint or tsconfig or something like that to be able to format Your your code and all of that context, and then you ask an answer, and there's limits to how much Text you can send a lot of these AI things. It's called token.

It's called token limits, and the token limits just keep getting larger and larger and larger So much that you can like soon you can provide an entire like the newest version of ChatGpt CFT 3.5, we can pass it an entire hour's worth of transcripts uncompressed, which is unreal that you can you can do that.

Eventually, we'll be able to pass it 8 episodes of of uncompressed transcripts and ask it questions. Right? So Cody seems to be a little bit different in Like you can upload documents to it and it knows about your entire your entire code base, which is cool because Sourcegraph Was like the OG, tool for exploring your codebase. This is a function. Where is it used? What are the arguments? All of that good stuff, kind of similar how to how the new GitHub code search works. Sourcegraph has been doing that for quite a while, so they're No strangers to understanding code bases. Mhmm. Yeah. Totally. So is this is this free paid? What's the

What's the deal? Is a paid product, but Cody AI is free.

And when you when you do look at that or think about it, You wonder if that's like a free forever or if it's free until they hook you.

You know, Free until they make it part of your daily life. Now Mhmm. I I would have a hard time thinking that this couldn't And that being a paid product because there are others that we'll talk about, like GitHub Labs that even do very similar things to this. So,

You know? It says 250 GPT 3.5 queries a month, which is, How does it know about your entire code base if it's just using GPT 3.5?

You know? They do some scanning of your code base, but I don't know how they're sending that information then to

g p t as kind of context. Right? They're just really good at the context. So, forget what I was saying about not having to provide context. They still do that, But they must be very good at it. Yeah. Like, how could it upload your entire they must be able to summarize Your entire code base very succinctly, and maybe they had access to the larger

token limits. Yeah. It's interesting.

But even just even Even the noncontextually aware tools I found to be really nice. Very good. Like I said, just just the code smells and the optimize this code, I found it to be a neat little thing. Sweet.

Next one we have here is, Amazon Code Whisperer. I'm gonna I'll talk about that one because I did a video on it. You can go to my YouTube and take a look. So Amazon has released their own version of GitHub Copilot, and it's pretty good.

At least when I used it, it was probably like 90% as good as Copilot.

And it's really hard to It's really hard to, like, say these things definitively because they're constantly being updated and constantly being retrained.

So If you want to try something, especially because CodeWhisperer is free, who knows if it will be for a while? But GitHub Copilot is not free. So if you want something that's just as good, in my opinion, you definitely should check that one out as well.

Interesting. Yeah. I've never used this.

I don't know why. There's just too many of these tools. I wanted to actually bring attention to the Copilot Labs extension for Versus code is a, a free extension for Versus code that's separate from the Copilot one called Copilot Labs, Which is really just like experiments.

And very similarly to this past one that we've mentioned, the source graph code.

Copilot Labs also has some interesting they call them brushes where the other one called them recipes.

So they have different brushes and and then they don't exist for everything. Like, I don't see Svelte as an option here for brushes.

But there's some interesting stuff you can do. You can select debug, fix Bug, add types, make it readable, document.

You can have some custom ones, list steps, clean.

So a lot of these little features here would be, you know, you select your code, then you could say document this code, and then it's going to, well, you could actually well, it looks like you can apply

several different brushes at once. You can you pretty much type in whatever you want. So I use the custom brush quite a bit. You do? Or I just select a block of code and I will say custom brush. And then I say, just like convert this to X, Y and Z or make this if statement a switch statement. I'm trying to remember. On Friday I did another one, I probably use the brush custom brush, like once every, I don't know, at least twice, 2 or 3 times a day, Which is pretty handy. I guess once the Copilot chat hits, I'll probably start using that a little bit more. It's only available in Versus Code nightly, and I don't use nightly as my daily driver, so I'm kind of waiting for that to hit Stephen, probably next month whenever those features land in stable.

Yeah. I'm curious here. I I've found that The in in the labs, occasionally, when I'm selecting things, it it gets my code occasionally a little Goofy in terms of, like, deleting things or or getting it wonky. So I did it's definitely experimental and maybe not one that you can rely on. But as far as, like, Just adding comments to code and stuff like that, it it works really well. And in general, I found it to be just a nice little addition to Versus Code. It. It's always good to have these tools around, like, if you have something that you you're having trouble typing or, like, I mean, like TypeScript typing or something that, In general, you just don't want to spend the time documenting or writing tests. Some of these tools can help there. Yeah. For my talk on AI encoding, I

tested out a bunch of them, And I found that the even the next version, which is chat, is even better than like the brushes is just sure It's just the same as like pasting it into Chat GPT and then saying do this with this code. It's just a different UI, right? But I find that Chat Is a bit better. You can you can reference Oh, really? You select some code like, I selected the following code. Let me know x, y, and z. Or here's a function.

Here's a bunch of tests.

Write the code that matches all the tests, and it's really good. I don't know if it's a bump up in In whatever model they're using, but I was pretty surprised. And that's just straight up Copilot chat, right? Yeah, yeah, it's it's a lot of these are invite only. So you have to just go to the GitHub next website and apply for all of them. And then usually for me, A month or 2 later, I'll get an email saying you're approved. So, like, I'm not getting any special treatment or anything even though we've we've had the folks on. What about, have you have you explored with any of these? So here's to lint rule.com

and an open commit are 2 interesting kind of outside of the outside of your application development flow tool. So, Lint rule is an LLM to review your code.

A command line tool, basically for code reviews.

And so I would imagine that it's using your config files, because it does say it's able to find things like code styles that are incorrect.

If somebody's doing you're you're doing a code review on some code that somebody has committed. Yeah. It it looks really interesting to me. I haven't I haven't used this yet. It's been on one of my my things. Get started for free. No credit card required. I would imagine that it's a paid product here, But definitely something if you're doing a lot of code reviews. Okay. $1 per 1,000 lines of code change. It's free. It's expensive.

Yeah.

Not not a non non factor there. The whole idea with the slint rule and I actually tweeted the author to give us a couple more examples.

Having a hard time, like understanding where this sits, and the idea is that it catches things that Rigid linting rules will not catch.

So like, let's say console logging users a user's name Like that's potentially violating PAPIDA or any other acts, right? And it's able to To know like your Linting World doesn't know what's personal information, but the AI can kind of say, Oh, looks like you're logging somebody's someone's username or Someone's password you accidentally let a log go through to production, in that kind of thing or.

I guess, like like database queries is another example they have, but I I would imagine but, like, what is it? What's the thing we're using for the syntax website? The Prisma, like your ORM. Yeah, Prisma yells at you whenever you do something potentially bad, right? There's a whole linting rule in there.

So he didn't he never applied. I don't know if he didn't see it's tweet or not, but I would love like a bunch of more examples of things that this could actually hit.

But it seems to be

kinda interesting. Yeah. And it does say, like, part of it is that you can write your linting rules in plain English.

So you could say like, yeah. What would you use what in that situation that I guess you couldn't do in a straight up normal config file? Maybe you could say, like, When I am logging

a table that has products in it, Only put 3 items on a row or, like, the one annoying thing to me is like a formatter is when you're trying to format an array and it, like, puts One line per thing, and there's not really any knobs you can turn on how to format arrays.

I don't know that this will A linter would catch that, but I'm curious if we'll ever get a like a prettier with AI. It seems a bit weird because The formatting is so rigid. Right? I don't know if you put AI into that. Yeah. Formatting is straight if this, then that. Yeah. But this could be like a multi step if this, then that, you know. But I like I have certainly hit lots of issues in my format or where I'm like, well, Like, the right one is what kills me. If I have an array of a 1000 things or a 100 items, I don't know. Let me put 5 on a row, you know, and that's not an option in Prettier.

It's either all in 1 line or 1 per line, and there's no in between. Interesting. What what about this open commit? Open commit is essentially a tool for For allowing you to generate,

commit messages. Yeah, it's a great idea for, for Git, right? Because GitHub It also has GitHub Copilot for pull requests. Kind of the same idea in that It will look at what you have changed and provide a good summary of what you've done. That's a perfect use case for AI because once you're done Doing all the code. Now you got to spend a whole bunch of time figuring out what did you actually do and how do I Nicely format this and put it into a big list. And it's sometimes you're just like, I don't know, bug fixes, you know, merge a thing. If you're like me, at that point, you're you're ready to be done. At that point, you're like, alright.

I'm I'm done. I don't wanna write a a long commit message, but those are important for, you know, keeping everybody In the loop as to what's actually going on. And I I think this thing's really neat. So, I mean, what the way they show you integrating it in their own docs Is to essentially put it into a GitHub action that will modify your commit messages.

So you you push up, you know, commit messages like fix the bug, And it will change it into chore, adds validation for if hook, exists to improve. And it even does like emojis for the different ones. So you could have, You know, emoji based. Yeah. Commit messages if you so choose.

Yeah. This this is great, and it uses your, Your own OpenAI key. So you're paying it through OpenAI?

And like, honestly, you're not you're gonna be paying pennies a month For this type totally because unless you're writing like gobs and gobs of code, it's not going to cost all that much to be able to parse The diff changes.

So it's I think you could you could add this and probably pay a couple bucks a month and no problem there. Yeah, totally.

Yeah. It's a it's a neat tool. Maybe this is something that we should use because I do feel like I can get lazy with it. I don't get bad with it, but I feel like I can get lazy with it. Remember we tried to to switch to standard commit? Conventional commits. Conventional

conventional commit.

Did you, did that ever

Take up for you? I I probably had to have it for, like, 6 months. And then, you know, the moment that the going gets A little bit more stressful. Yeah. Breeze right by in.

I always see people using it and then they can generate these beautiful things like change logs from the commits. Yeah. But I I just can never get into it. And this I wonder if this can do Conventional config. Yeah. It says fix feat. Yeah. Yeah. Like, you'd be able to then compile these instant changes. Yeah. Let's throw this On the new syntax website, it'd be kind of fun.

I'll do it, do it to it. Although like renaming the commits,

that doesn't change the commit hash, right? Just the description. Correct. Yeah. That's fine. Yeah. I'm cool with that. I'll throw this on. Next one is Replit Ghostwriter.

So Replit's doing, like, A bunch of really interesting stuff in the space. So obviously they are browser based IDE, but they are also Been training their own, LLM, large language. They've been basically making their own model and open sourcing it, which is Really cool. So they have a Copilot like extension called Ghostwriter for REBELLET That will autocomplete it and whatnot, but they also have their entire model available to you. You can you can run it on online. You can use the hugging face, which is a big hugging face is like a website that has models and data sets and What are called spaces, which is if you take models and data sets and put them together into little demos, that's what a space is.

And it's I went down the rabbit hole as well from my talk and going down looking into all the different Hugging Face models. There's some pretty cool stuff out there that is just

You can run it yourself. Yeah. Speaking of hugging face models, one of the ones on here, StarCoder, is exactly that. It's just a model. It's a It's a language model trained on, source code and natural language text, incorporates more than 80 different programming languages. And From what I've read, this is supposed to be like the most advanced, language model for programming languages.

But when I use this extension, I don't know if it's just me, but I I kinda am unable to get really useful code out of it The same way that I am out of Copilot Yeah. In terms of JavaScript and Svelte.

And I don't know. Maybe I'm not telling it The right thing. Because it seems like a lot of times the code it's spitting out to me isn't like, it it spat out, for instance, an HTML form in Svelte. That's fine. But then it also gave me some JavaScript that was not JavaScript.

It looked like another language. So I I don't know what's going on there, but this one every time I read about programming language models, Starcoder is listed as The one to try. So, yeah, I'm sure I'm sure I'm doing something wrong here, but it's something to keep an eye on, at least. I found the same thing as well. I had Tested it out for my talk, and

I was like, oh, this is like pretty good, but it's it's definitely not at copilot levels.

But I think the other part is that they I think it's supposed to be much smaller. It's something that you can run locally or on your computer.

So certainly one to keep an eye on, but certainly not. At least from what I've tried, not the best.

There is Code Geeks, which is again another kind of open source CodePilot Alternatives with an X, by the way. Yeah. X, which makes it super cool. It's codegeeks with an X dotcn, Which is I think it must come from China, then.

I see they have Chinese translation on there. You can get the Versus Code extension for it.

Have you tried this one yet? Yeah. Yeah, I.

Again, it's it's like it's kind of funny that there's so many and it's it's kind of hard To subjectively see if it's good or not. I guess that's why. So OpenAI introduced a bunch of Like baseline tests, almost like unit tests.

Mhmm. Because people start saying, OpenAI is not as good anymore.

You know, and and like, how do you other than people just saying, Yeah, it's it's not that good anymore. I've been noticing it too.

How do you measure that, right? So OpenAI now has a bunch of what it's called. I'm pretty sure it's the Evals repo that they have.

So theoretically, I guess you could you could run a massive set of evaluations against all these different frameworks and see Which one is subjectively best? But what I found that's why I did my I tried to write the exact same application line by line In both Amazon CodeWhisperer and in Copilot, and I just kind of did it 1 by 1. And then you could you could actually see The difference between the 2 and say, Oh, that one didn't catch that are both of them understood to use asyncawait instead of just promises, and they all did pretty good. So you almost need to, like, do those line by line and put them into a vowels and then run it against all of these LOL, that's out

there. How did you use Code Whisperer in Versus Code?

Just install the extension.

You have to Sign in to Amazon.

Is it just the AWS toolkit extension?

Yeah. In typical AWS fashion. It's packed. Yeah, they make you get all of the other stuff. Yeah.

But it's it's well worth Worth doing it. Next one is, Codeum AI,

and this is a modern coding.

Extension again, it can work with Versus Code. There is an enterprise plan. It also works with JetBrains stuff and Jupyter Notebooks and a bunch of other stuff.

But, again, it it's, in house models, not an API wrapper.

Like, in and, Again, it's just a it's an AI tool that helps you do stuff like generate unit tests, infuse AI into your coding. Basically, It's doing all the same kind of stuff that these other ones are doing.

And and some of them is really you write comments, you get tests, or you write comments, you get code. You write code, you can get Comments you can write code. You can get tests. You can maybe clean it up or or ask a different thing. There is a chat. There's a playground. Just another one of these tools. And, this one is at least interesting because it is you know, you get things like rapid auto complete like you do with Copilot, You know, which is one of the things that I I Like, this one's local. Like, you can run it on your computer? Yes. It says in house model. Well In house. So so so their own, CodeM's own, not an API wrapper, aka this thing isn't just calling a text string to OpenAI APIs.

Okay, because I'm curious what the like, maybe people who listen to this, if you work for a company, Are you allowed to use GitHub Copilot? I'm sure there's so many companies that say, absolutely not. You're not Uploading snippets of our code base to some random AI thing that's like untested for whether it's going to be secure or not.

So I'm curious when we'll start to see, like, in house versions of things where you're like, oh, yeah, I can use Copilot, but we have to have it running

On our server and I have to be on the VPN to to use it. Okay. So they do have a nice little Copilot verse Codium. Okay. So they're saying that Codium is Free rather than a paid product, functionality.

You get CodeGen, like on both of them.

You get an in ID integrated chat and search, but you get those in Copilot too. This one works, it seems like, with more editors. So instead of just Versus Code and JetBrains, it's also with Jupyter. It's also with, Emacs.

And it works on all of the web IDEs for forget or for Versus code. It works with way more languages apparently.

And let's see. Up you can opt out of code snippet telemetry. Never never train generative model on private code, And it looks like they have lower latency.

So it does seem like they're kind of saying, alright, this is very similar to Copilot, but it's free. It's lower latency, And we're gonna give you some nice things with it. They also do a side by side with CodeWhisperer as well.

And they said looks like according to this, CodeWhisperer has, like, a really bad latency score. Oh, yeah. You know, I really noticed the latency when I was

doing Dev on an airplane Because the Internet was fast, the latency was long. Like, if I do a Ping, if I run a speed test right now, I'm I'm at a wired connection in the city and I get a let's see what my ping is here.

3 sorry, 10 millisecond ping, Right. Like, that's that's decent. But when I go out to the cottage, I'm on Starlink. Often the ping is 100 150 milliseconds, you know? And I definitely noticed on the airplane. I'm pretty sure I noticed it on Starlink, but I'm going up there tomorrow, so I'll be sure to note if that's an issue because You think, like, 100 millisecond delay? That's significant in something that is just part of your regular Devflow. Yeah, totally.

We also have Tab 9. They're a sponsor of us, and there was also Kite. Kite actually shut down a couple of months ago. Yeah. This was really early. Too early. Yeah.

They they were really, really early. Like they ran this was like years ago. I'm pretty sure I talked about it. See, this is This is where we need the syntax search that I was working on, which is like, I want to be able to say, hey, in what episode did we talk about Kite? Because I'm pretty sure that was Couple of years ago, so closed down. Tab 9 seems to be.

Around It's it knows on all of yours, but, I don't know. I haven't I haven't heard too too much about from them lately.

I think they'll they'll go with a more corporate on prem solution.

You think? Yeah. There needs to be something like that that just

Fills that void. Oh, yeah. I just I just loaded their web page. AI assistant that speeds up delivery and keeps your code safe.

Private, secure, and compliant adapts your code base. Yeah. Look at you. Wow. They are gonna be the, like, corporate version of this type of thing. They're always a cool company. They were, like, one of the first it in the space of this AI finishing. Well, because when when those first we started to get people talking to us about them before,

the AI anything was stuff people were talking about, and it did always feel like kind of like Okay. Sure. What can this actually do? But it almost seems like a a switch was flipped in my brain to say, oh, no. Wait. The these things I've probably always been this cool, and I'm just skeptical about it. Exactly.

Yeah. So let's actually another thing I, you know, I I think before we can Wrap it up here is that just having, like, a chat GPT window, whether you're using it through Raycast or through chat GPT, I know a lot of these tools might just be like a wrapper for that as well. But I I like having just a chat interface and giving giving it prompts. You can really get a lot accomplished out of it, especially You need some dummy data. You need the format data. You need to give me this in the form of this.

You know, it will it will do all of those things as long as you can craft your Responses. And one thing that I always do when I'm working in a, like, a chain, and I'm telling it about my code base and I say, Hey. All prompts need to be answered with 1 using snake case, 2 using this version of node. We're using these libraries or whatever.

And, typically, the code it gives back to you from the point forward will be meeting all of your requirements. Because at one point, I I remember asking it for something, and it gave me, like, an older version of Node. And I said, I believe you can do this with the crypto package. And I was like, oh, right. You are, but you have to use this version of Node. I'm saying, Well, we're always using that version of it, so let's go. Yeah. Yeah. It's nice when you don't have to give context. Can I can I read you a

Sloppy prompt I gave it the other day in the data I came back with? Yeah, so I said This is to GitHub Copilot chat. It's giving you an array of people. Each person has a first, last, and age property. Each person also has a nested array of 1 to 3 jobs, Which have a start date, optional date description, and a manager who has the same properties as a person generate types. So, like, that was the laziest, most rambley Generate me a bunch of dummy data. I intentionally was rambling on it right there and I tell you it legit.

I tell you, I tell you, give me the person type.

Give me a job type and then it correctly relation to those Like 1 person has many jobs.

Each job has a manager that could also be a person.

And the manager can have a job, you know, Lee nested and it nailed it absolutely nailed it with the best dummy data.

And I, as someone who creates a lot of tutorials and A lot of dummy data. Mhmm. And I have used Scott and Scott's dogs as so many examples.

You know? I do the same thing. More and more better examples.

Right. I hear you.

So that's really it for AI tools. Obviously, There's more in this space all the time, and I'm sure there's many that we missed because there's just so many things in this world.

But, You know, the thing that I I think is the big takeaway for many of this stuff is these tools will help you accomplish tasks, But they're not necessarily going to do the task for you. And they're not going to they're not going to always write Wallace code that you can just use without understanding.

So the biggest thing to keep in mind that I would say at the The end of the day is to always know if what these things are outputting is correct.

Don't use them To ship code without knowing what you're doing because I found many many instances where the code is garbage.

Mhmm. But If you know what you're looking for and you can tell at a glance, it did it. It got it. This is right. Whatever. Write me a regex that does this. Okay. It's right. Write me some tests or some descriptions. Alright. Do that. It can save you so much time.

And having these tools at your disposal, You know, it it takes some of the manual effort part out of writing code that can just be a pain sometimes. So, yeah, If there's any specific AI tools that you're using, that you love, that you can't put down, let us know. We will talk about them in another episode. I think this is a space that is going to Just continually evolve. Alright. Thanks, everybody, for tuning in. We will catch you on Wednesday.

Peace. Peace.

Head on over to syntax s.f m for a full archive of all of our shows.

And don't forget to subscribe in your podcast player or drop a review if you like this show.