AI's an amazing enabler for devs

00:00:00: When talking about AI from a developer's

00:00:03: perspective, I think there's one aspect that's a bit

00:00:07: Because of course, AI has a big impact and of

00:00:11: course, you can look at AI and dismiss it and say it's,

00:00:15: uh, all garbage, it produces horrible, uh,

00:00:18: results. You could say it will replace

00:00:21: developers and we'll all vibe code our apps in

00:00:25: the future. Or you could have a more nuanced

00:00:29: take and say what I believe, "AI is, is a great

00:00:33: assistant, a great tool that can make you productive,

00:00:37: more things done if..." And that's the important part, "If

00:00:41: you have some experience as a developer, if you can combine your knowledge

00:00:45: with AI." And I've said that in many v- videos and episodes and

00:00:49: I don't wanna repeat all of that here, but there's one thing

00:00:52: that's missing maybe or one thing that I also find important

00:00:56: when we look at AI from a developer's perspective.

00:00:59: And that is the fact that AI, even

00:01:03: if we are an experienced developer, enables us to build

00:01:07: things we previously weren't able to build.

00:01:10: And also important, it can help us

00:01:14: learn new things and expand our skill

00:01:17: set. Now I will say there also is the

00:01:20: danger of going the opposite direction.

00:01:22: There is the danger, and I've talked about this too, of

00:01:27: relying on AI too much, of not understanding the code it produces,

00:01:31: maybe of not really looking at it and of ending up in a code base

00:01:35: where you don't know the code, where you don't understand what's going

00:01:38: on. Uh, that's definitely not the direction you

00:01:41: But let's say I wanna build my own web-based

00:01:45: image editor, something trivial where I maybe can just

00:01:49: the saturation of an image and save that as a new

00:01:53: image. Or I wanna build my own very basic video

00:01:57: tool that, uh, encodes a video, maybe extracts

00:02:01: the audio from a video. I wanna build my own tool because I wanna

00:02:05: an overall workflow or anything like that.

00:02:07: Now previously, I'll be honest, these are things that would've

00:02:11: taken me quite some time to build because I had basically no

00:02:15: experience with working with the Canvas API or

00:02:19: of image manipulation in the browser.

00:02:21: I had no experience with MPEG. So these are things where I

00:02:25: would have to dive into the documentation, read tutorials, watch videos.

00:02:29: And whilst that all is fun and builds up some knowledge and

00:02:33: is therefore in theory important, it's also

00:02:37: something where I probably wouldn't have had the time so I wouldn't have

00:02:41: done it and I would've found other solutions to my problem, so

00:02:45: everything's good to some extent.

00:02:48: But I do believe that it's now actually a

00:02:52: better time to be a developer because we can

00:02:55: now build any kind of utility tool, any kind

00:02:59: of workflow we wanna build, even if it involves

00:03:03: things or skills that we don't have

00:03:06: yet. Because if I need a, a tool that

00:03:10: extracts audio from a video, I can just prompt AI

00:03:14: to build it for me and I very likely get a tool

00:03:18: that works. Is it perfect? No. It will probably not cover all

00:03:22: possible edge cases, it will probably be buggy to

00:03:26: some extent, but it might be good enough for me.

00:03:29: It might be good enough for me to get the job done which I want to get done.

00:03:32: If it fails on a four-hour video but works on a ten-minute

00:03:36: video and I only have ten-minute videos, then this is

00:03:40: not a bug or a problem I need to tackle.

00:03:43: So I don't need the perfect tool, I just need a tool that's good

00:03:47: enough. But isn't that vibe coding?

00:03:49: Well, to some degree it is, but here's the interesting thing,

00:03:53: me. I really enjoy the process of building

00:03:56: utility tools, little helpers that I can

00:04:00: use and then I enjoy the part of diving into the code

00:04:04: and doing some research of what's happening there.

00:04:07: But since I already got something working, since I already have

00:04:10: that initial solution, I don't have

00:04:14: to dive into the docs and try to find out which parts are

00:04:18: important for me. Instead, I can just read more about

00:04:22: the, the methods, the tools, the, the functions that are

00:04:26: being used in my code and understand what exactly

00:04:29: And that allows me to build up an understanding of how that

00:04:33: technology, that library, or whatever is getting used in my code

00:04:37: works under the hood. And I've learned a ton about working with

00:04:41: Canvas API, about web animations, about tools like

00:04:44: MPEG, and many other things just because I built

00:04:48: own little helpers. And these are not just fun projects, by the way.

00:04:52: These are actual things that then in the end might find their way

00:04:56: into an actual product I'm building.

00:04:59: Obviously then polished and with, um,

00:05:03: with more knowledge from my side then because of course once I build something

00:05:07: production for other people to use, I 100% want to make

00:05:11: sure that there are no bugs in there, or at least no

00:05:14: bugs, uh, I, I know about and I wanna make sure that it's stable,

00:05:18: that it's fast, that I use common best practices.

00:05:21: So of course I will then take that time to polish it and

00:05:25: to learn more about it. But for getting started with the learning

00:05:28: process or if we're doing some one-off tasks, I think we're

00:05:32: living in the best time possible because AI is a big

00:05:36: enabler there. By the way, also for example, if we're talking

00:05:40: about technologies or programming languages we're not

00:05:44: that good in. Now, for example, I'm way better in

00:05:47: JavaScript than I am in Python also because I

00:05:51: haven't written a lot of Python code for the last three to four years.

00:05:55: Lots of small utility scripts and workflows,

00:05:58: definitely, but nothing complex.

00:06:01: So I don't have any expert or deep expert

00:06:05: level knowledge there. Something I wanna change again,

00:06:08: But it is what it is. Here of course, AI can also

00:06:12: help me get stuff done and then either I

00:06:16: remember the syntax I see and I remember what's happening there or I

00:06:19: can then again dive deeper. And I think that's a

00:06:23: very valuable thing or advantage we get out of

00:06:27: AI, because it allows us to do stuff that otherwise would've been

00:06:31: impossible or maybe just not worth our time

00:06:34: and also grow our skill base. I

00:06:38: think that is the, the important part

00:06:42: pure vibe coding where you don't even look at the code, where you don't care about

00:06:45: the code. I care. I like learning. And as I've said in other

00:06:49: episodes as well, I can only recommend going

00:06:52: deep, uh, as a developer. You want the deep

00:06:56: knowledge, you wanna be better than AI, you wanna be an

00:07:00: expert in certain matters and you wanna understand what AI is

00:07:04: doing and always get to a level where you can correct AI or

00:07:08: steer AI. Because as long as that's the case,

00:07:12: you'll have, uh, a reason to get employed or

00:07:16: whatever your goals are. You don't wanna be on the same level as

00:07:19: You wanna be above it at least in certain areas.

00:07:22: Obviously with the broad knowledge AI has, you can't beat it

00:07:26: everywhere but you can beat it in certain domains and it can be a

00:07:29: great enabler.