Fancy Wednesdays And Single-Purpose AI Tools

Transcript is auto-generated and may contain errors.

Hey folks, I'm just headed to work here. It is Wednesday. It's apparently fancy Wednesdays starting this Wednesday. So, I talked with a PM last week at work and we were talking about wardrobes and such and I had said I literally spent and this is way before code commute, but I spent a like a year transitioning my wardrobe to not just being like jeans and t-shirts. Um, which for anyone who's tried doing that, it's a super awkward thing because you start trying to wear nice clothes or nicer clothes cuz you're changing your wardrobe and all of a sudden you got people like the first day someone's like oh job interview and you're like ah like just trying to just trying to wear nicer clothes and then it happens for like a month and then after that like no one cares. Um, but yeah, like it took me a long time to kind of transition over to like what I would consider nicer clothes and I was, you know, I felt good wearing them.

Uh, finally had like fitted shirts and stuff. I'm I'm real short, but I'm kind of wide because I was doing a lot of bodybuilding, so like nothing fits. Um, so I had like tailored clothes and stuff. And then we had the lockdown and I was like, I'm not wearing stupid fancy clothes if I'm sitting at home on a call with my my camera off, right? Just felt kind of weird to try getting like more dressed up to like uh to go nowhere. So, uh, after doing all that work to transition to wardrobe, I, uh, kind of regressed to not ever wearing nice clothes, which sucks because I I like wearing nice clothes. So, we were talking about it and we said Wednesdays we're going to start doing fancy Wednesdays. So, I got fancy shoes on. You can't see them cuz I'm in a car. And I have a terrible feeling he's going to either not show up to work today or going to show up and be like, "Oh yeah, I forgot." Um, that's okay.

So, the topic for today's code commute, we're going to experience dev subreddit. Uh, we're going to make this one a little bit more generic because I feel like the post is kind of generic, but it's kind of like is everyone getting sick of singleuse AI tools? kind of feels like, you know, everything there's AI in everything and I don't know, I get it. Um, you know, every company's trying to put AI into whatever they got. It feels like there's a special use AI tool for this and for that. And so I wanted to talk about this from the also from the perspective of companies that are trying to to transition to you know how do we start leveraging AI more and more in the things we're doing whether that's building features out or you know like internally right like um you know the the common

go-to thing is a chatbot right like that's the first we see chat GPT we're like we could have our own chat GPT Um, and then like I want a specialbuilt chat GPT that does this and someone else is like I want a specialbuilt chat GBT kind of thing that does this. And we get a lot of individual things. So I wanted to talk about this. I think it's kind of uh interesting. I think there's some I think there's just like a lot of growing pains and stuff in the industry right now. And we're seeing the symptoms of this in so many ways, right? like we see uh obviously there's a super difficult job market. We see uh what to me kind of feels like companies like everyone's trying to rush to put AI into everything even when you're like dude we really don't need AI there.

Like that's not actually solving a problem. But it's like if if we don't do it then someone else will have done it and then we've missed it. I don't know if that's like a good reason but like that's what it seems like. That's what it feels like. Um, so I I just feel like there's a lot going on and like we have uh internally teams are like, "Well, I got to build like we need to use AI too. Like we should have a chatbot. We should have a a thing that does this. We should have an agent that does that." Um, let's go build it. Someone else is like, "We're building one." And you get, you know, a million different little things going on. And it's, uh, like I said, I feel like part of this is kind of growing pains for our industry.

there's like with AI things are moving so fast uh in this direction and I I don't know like in my like I don't have like a 40-year career um kind of thing but it's like professionally like 15 years writing code for like just over two decades and um I don't like I've seen some things where it's like oh like things get picked up pretty quick and like it's trendy but it seems like it's in niches. is and uh or you know like everyone like everyone things got to be a micro service and then we transition to like the complete opposite way but these were things that seem like they happen over a while and I don't know maybe that's just in hindsight it seemed like it was a while maybe they did pick up super fast and then everyone tried to move away super fast it just to me it seemed like it was over quite a while and with AI it feels like it's literally everyone everywhere at warp speed.

So, uh you know, I think they're they're comparable, but the speed at which we're going through this stuff is kind of nuts. Um and then as a result, the growing pains like really stand out. So, um give me one sec. I just want to have some water here cuz I'm about to die. Okay. Less dying. Some caffeine in that, too. Oh, boy. Um, so the I I I think a couple angles I want to touch on from here. I think from like mostly probably from like developer tooling and whether that's what's available to us andor what we're trying to build internally as development teams. I'm going to start with what's available to us because I I don't know. I feel like this is probably more relevant for a lot of people, but we have so many things available to us, right? there's, you know, a million different people and businesses trying to put AI developer tools of some kind in front of people.

And um it's interesting, right? I I think one of the comments I saw on this Reddit post that was kind of like it's kind of funny, but I I think it's I also don't disagree with it. It's like, are you actually forced to use all these individual tools? Like is that what's happening or is it like you kind of got caught up in it and now you're like oh crap I have like you know 50 different little AI tools I use for all these things because I don't know at least for my own development outside of work. Yeah I use lots of various things but I don't feel like I'm forced into any of them if that makes sense. like uh and this is one of the meta points I want to talk through here is like the fact that we do have options and I think that is kind of nice and I don't think that we're necessarily forced into using them that might be different for you at work though.

So just uh you know caveat there but like for us that aren't locked into certain things like is this really a pressure or a a constraint that we've accidentally put on ourselves? It's the second time my camera's died today and it's literally plugged in. So, I'm uh probably going to send an angry email to Sony or something. But um I love this camera, but the battery is like the biggest pile of hot dog I have ever seen on any device. It's outrageous. Completely outrageous. Um what I was saying if I can remember was around like uh AI tools and like uh you know for for those of us that aren't locked into specific things because of work got lots of options right so I you know I have a handful of things that I like cycling between and I find that as things are advancing what what I'm noticing is that I'll gravitate towards something because I'm like, "Oh, now this thing does it like awesome.

I'm going to use it for for this reason." And that's like a reason I stick to it. And then what happens is like the other tools kind of like converge on that. And then what happens is like I seem or I feel some underperformance in one of the tools I'm using. Then I'm like, "Okay, I got to go switch to the other ones." Like I can't stand using this. And I'll kind of give you a bit of an example here. Like I uh I really like using GitHub Copilot. And I know there's going to be people that are like, "Oh, it's because you work at Microsoft." Like, nope. I I like working with GitHub Copilot. Uh because it's super convenient for me to go sit in bed, make GitHub issues, and assign them. Um, I know, yes, I can do this with cursor now. But that's kind of my point with this little bit of a story here is like, yes, I can do that with cursor now.

I can go talk to cursor and say, go work on this and it will go make a PR. And I couldn't do that before. And I think, if I'm not mistaken, I think Claude has just launched something like this. But I couldn't use Claude that way. Claude was great if I was sitting at my desk. And either I wanted to watch AI build things and kind of do it interactively with AI or I had a really awesome plan and I wasn't planning to be at my computer and I could just let it go off and do some and then come back and check on it. But I like cycling through um GitHub Copilot. I do use Cursor. I do use Claude. But like I have periods of time where one of those is more of my primary tool than the others.

I've had times where it seems like no matter what model I'm picking in uh in Copilot, like even in Visual Studio, I'm like, I don't know how this thing could get any more stupid. Like it's infuriating. And you know, I'll switch models and I'm like, okay, this one's a little better, but it's still being stupid. like and I'm not even doing anything more complex than the previous thing, which it did perfectly fine, but I get frustrated enough where I'm like, I'm just I'm simply losing time now. Let me like I'll literally sometimes go to chat GPT and be like, walk me through this here. Like, what would you do in this case? Or like how would we refactor, redesign this code? And it does exactly what I'm hoping to see. And I'm like, great, copy, paste, done. Um, so like even something like chat GPT, which I'm like it doesn't have any context of my codebase, whatever.

It has the least amount of information of all these tools and it just does a better job, right? So sometimes I'll get frustrated with C-Pilot. I'll do a quick thing with chat GPT and then I'm like, okay, now for the rest of this feature development, I might go to cursor now, right? Or I'm going to go to Claude. Uh, suck it up and live in my command prompt. I'm not I don't love working in a command prompt. That feels like pass backwards to me. Uh I know there's a lot of people that love working in command prompts. Their IDE is a command prompt. You know, um they somehow, you know, all of their skill and experience is measured by how much they are superior in a command prompt. And I'm like, I just I don't I'm happy for you, but that's not uh that doesn't feel ideal to me.

And because you're not programming for me, I have to program for myself. You know, your your points around how much you love a command prompt are kind of moot when it comes to my own development. So, um, but yeah, there's times where like I'll go to Claude and hang out in a command prompt and we get done. And then there's times where I'm like, "Okay, now I'm super frustrated with Claude or I'm like, I have a handful of things I want to work on or I want my computer back so I can go work on stuff myself and I need co-pilot again. I need GitHub C-Pilot to go work on issues or I'm going to fire them off with cursor agents." So, I like having this flexibility to go between these things, right? And I personally don't find myself caught up in like like micro tools that go do this stuff.

It's like really four things. Five if you like separate out C-pilot and Visual Studio from GitHub Copilot. So GitHub Copilot, Copilot and Visual Studio, Cursor, Claude, and good old chat GBT all cycle between those five things. And uh I don't think I can't remember if I released the video on Dev Leader yet, but if it's not out yet, it's coming out very soon. By the time you watch this, it might be out. Um so on my main YouTube channel, I I'm trying to give some examples of how I work through these tools differently. Not because one is automatically better or worse than others, but they're just different. But um yeah, I don't know. Like maybe maybe some people are like well we're using a special like what like what is it code rabbit or some for like code reviews and like like is augment code something?

I don't know. There there are like a million things but like do you need those? Do you need all those things or like is this just something that you kind of did you do this to yourself? I don't know. um you know there's a trillion little MCP servers because now everyone needs to have MCP like but do you need that or is this just one of those kind of like little traps you fell into where it was shiny and you chased it and now you're like now I have all these things and I don't know like how do I start living without them? Um but that's kind of my take on it at least for for development outside of work. If you're at work and you're being forced into this kind of stuff, I would say this is probably one of those opportunities to have like, you know, raise it with team discussions.

Do you do retrospectives? Do you have a forum where you can raise concerns on this kind of thing? Because if tools are being integrated, like what's the point of a tool, right? Let's get a little bit philosophical. Point of a tool is to help us, right? It's to make some problem that we have easier. That's what a tool is. So if you have tools and they're not making things easier and you're working against the tools, that does not sound like the goal of a tool. Now, you might be the only one, in which case might be worth talking to other people on the team to see how you can get more experience with it, understand it better, right? Like that. And no, no shame in that, right? All this stuff's moving so fast. It's moving so fast. If you're like, "Dude, I just like everyone else is prompt wizards around me and I seemingly suck at prompts.

I can't get AI to build for me." Okay, like that. Like, don't don't panic. It's one of those things where like if other people are able to do it, what can you learn from them? Is it just so just so happens to be that the things you're trying to do are really odd and complex or is it truly that you're you know maybe you're lacking some of the understanding in in how you need to to work with these things? Like it truly is skill issues, bro. And if if that's the case, like again, it's okay. If you have a skill issue or you need more experience with something, what do we do? We spend time practicing it. And that's okay, right? So like don't don't feel shame in this stuff. That's fine. Like I will admit like I am not a great prompter. I'm not. And there's sometimes in hindsight where I've asked AI to do something and I'm like how stupid could you possibly be?

And then I like see what it did and I see my prompt and I'm like okay. Like I didn't actually tell you explicitly what I wanted. So like yeah maybe maybe that's my bad, right? Um, so it's going to be a learning process for a while. And we haven't been doing this for a long time. I've been writing code for over 20 years. How long have I been prompting agents? I don't I don't even prompt humans the same way that I prompt agents. I've been managing engineering teams for 13 years. I don't talk to people on my team like I talk to an agent. With an agent, I have to be very, very explicit. And if I talked to a team member like that, I feel like I I probably would not have been a manager for this long because people would be like, "What the hell is this guy doing?" So like, don't beat yourself up for it.

It's fine. But like acknowledge that there's an opportunity where you can improve and then come up with a plan for how you want to do that. But I would certainly try to raise visibility. If you find that you have too many things at work that are AI tools and integrations and like it's becoming messy, like raise awareness. Talk to people on the team. Is that the common is that the consensus? Like what's going on? Like did this just happen because it was new shiny syndrome? In which case, cool. Like how do we how do we work that back to like the core helpful tools? How do we get some alignment? and go from there, right? But this is an iterative process. It's not you might make improvements doing this and then in two months given all the new shiny things like it might happen again. So this is just goes back to what I said around growing pains, right?

That everything is new and shiny. It's going to be some growing pains for a bit. So raise awareness and chip away at it. I'm going to shift gears a little bit here um because I wanted to talk about how this looks when we're building stuff at work and I don't want to talk about feature development necessarily although um you might be able to kind of frame this in a similar way but uh one of the things that I've noticed and again growing pains kind of thing is that we have situations where people are like oh we we need to use AI more in what we're doing like there's such an opportunity here and um I feel like the go-to thing is like everyone's like well like what's the de facto thing that we're using in AI? It's like well I need an agent for that. Just need an agent.

Give it to an agent. An agent can do it. Okay. Well, what's an agent in this context? I don't know. But like if we have chat, we can add an agent to the chat. Okay. So, if we're after AI, we all use chat. We can add agents to chat. Cool. Let's go build an agent that does this in the chat. Let's go build another agent that does this in the chat. Everyone is building custom one-off agent things. And I don't necessarily think that there's something wrong with this. In fact, I think it's really cool that people are getting excited to go build things, right? That's that's what's really tricky about this kind of like thing is that I think it's really important that people are like we see that there's value in this. We see opportunity. We're curious. We want to try. We want to explore.

We want to do something with it. I don't want to shut that down. But what I get nervous about is when I see yet another chatbot and I go like what what like what what is net new here? Because like I'm assuming the logic you wrote I find a lot of the time it's like the logic is pretty straightforward for what they want some chat bots to do but we had to go spend a lot of time going to put together a chatbot or to integrate it or whatever. And in my mind I'm like, why are we paying that price over and over and over again? Or you put some logic here and only it's directly in your chatbot. So like nothing else can use this logic. So what I feel like we're experiencing um and I'm speaking from, you know, my experience, I don't know if this is the same for other people's organizations.

cuz I think it also depends on like your teams how much like how active people are with using AI right so yet all the developers are using like clawed co-pilot cursor something like that um like that's awesome are people building internal tools with AI that's kind of what I'm talking about here like kind of going beyond just what am I doing in my my IDE and Um, yeah. I think what I'm starting to feel like is like, okay, we're we're trying to expose capabilities to to agents and I'm trying to do a little bit more of a push to say like, look, we already have an agent in the chat. These other teams have agents over here. Like, what's the thing that we have in AI land that lets us give capabilities to agents? And those are going to be like MCP servers, right? We can offer tooling to existing agents so that LLMs can call in, leverage our tools and leverage that functionality that we're building.

So it's not like coupled specifically to a an agent somewhere. Does that mean that you like can't write a new agent? No. But I would say that if you're writing like interesting logic for something that the agent's going to do, I would highly recommend just figuring out a way to like pull that out into something that's, you know, MCP serveres so that it can be releveraged because I'm getting more and more nervous that we've seen chatbot after chatbot after chatbot and yes, they have been helpful. Yes, there's been interesting learnings there. I just feel like that part is diminishing returns and instead of continuing to invest super heavy into that, I would much rather see that we're we're going cool like what is the way that we can share some of this stuff. Um that's been sort of my most recent exposure and the reason I wanted to bring that into like this conversation just goes back to like many little tools, right?

I think when we have so many little tools like the problem is in my mind exacerbated when we're the ones building all the little tools. So instead of having more and more little singlepurpose tools and when I say tools here I mean it's kind of confusing nomenclature but I mean like uh agents or whatever right uh I would much rather that people are literally building like MCP tools and that way we have fewer um I don't know fewer agents I guess is maybe the way that I want to say that fewer agents or um you know tools as developers that we have to go use or or code for, right? We have we don't need to go, oh cool, it would be great if this agent could do this, but like oh we already have that functionality in this other agent. Okay, like let me copy and paste the code over and like we'll duplicate it.

It's like, well, why not pull the capability into a an MCP server and relever it? And I I don't see that shift happening yet, but I I don't know. I think growing pains again, right? My mouth is super dry today. Um, yeah. I don't know. I think that might be it for this topic. I I feel like I'm going to be starting to repeat myself, but I do like I'm I'm excited like while two of us wanted to switch lanes at the same time. Um, I'm excited about that stuff. while I see some like what looks like repetition to me or like duplication, I'm like it's kind like I feel like it's the same thing in in software, right? Like sometimes these things look like they're identical and then we're like, "Oh, we got to we got to follow dry, right? Like don't repeat yourself, so we got to make a common thing." And it's like, well, maybe they look similar, but they're not actually like fundamentally the same thing.

So, I don't know if we've truly seen enough examples of these things to to know. So, maybe maybe I'm completely wrong in saying like fewer agents, more MCP servers or MCP tools. I don't know. I could be wrong in saying that. my my observations now are telling me like that's the direction to move in, but we're early, right? Growing pains. So, anyway, I'd be curious to hear what other people have to say. Like, are you in a position where you're you're stuck using a bunch of different AI tools like and you don't have a way out? That's kind of how I interpreted this Reddit post. Maybe I just misinterpreted it. Um, do you find that like you want to use a bunch of different tools cuz that functionality is spread out everywhere? So, it's not forced to, it's just like things like there are really valuable singlepurpose tools scattered and you're like, I need this for that.

I need this other thing for this other thing. Um, I'd be curious to hear that. I don't personally have that issue. I rotate between a few core things which are really just like agentic coding tools. And then I honestly find like chat GPT is still like really really helpful for just like talking back and forth with for like brainstorming stuff. So that's kind of my like my core subset of things. Yeah. Would love to hear from other people. Friendly reminder that this channel is driven by your questions. So um please leave your questions below in the comments or you can go to codemute.com. You can submit questions anonymously that way. And then I I did give a reminder earlier, but I do have other YouTube channels. So, Dev Leader is my main YouTube channel. If you want to see some videos with coding with AI tools or you have uh you know requests or you want to see like tutorials on on some AI workflow for coding, like let me know, right?

I can uh you can ask here. I can go make the video on Dev Leader. All those videos are like uh they're edited. It's less less rambly from me, more around like we have a goal, let's go set out to do it or let's go explore something together. And uh I have an editor make that uh a little bit less boring than just me blabbing away here. So let me know if you're interested in that. Be happy to make some content over there for you. And I will see you in the next video. Take care.