WEBVTT

00:00:00.100 --> 00:00:03.540
When talking about AI from a developer's

00:00:03.600 --> 00:00:07.540
perspective,
I think there's one aspect that's a bit

00:00:07.580 --> 00:00:11.100
Because of course, AI has a big impact
and of

00:00:11.240 --> 00:00:15.220
course, you can look at AI
and dismiss it and say it's,

00:00:15.320 --> 00:00:18.750
uh, all garbage, it produces horrible, uh,

00:00:18.820 --> 00:00:21.980
results. You could say it will replace

00:00:21.990 --> 00:00:25.890
developers and we'll all vibe code our
apps in

00:00:25.900 --> 00:00:29.310
the future.
Or you could have a more nuanced

00:00:29.540 --> 00:00:33.300
take and say what I believe, "AI is,
is a great

00:00:33.309 --> 00:00:37.170
assistant, a great tool
that can make you productive,

00:00:37.280 --> 00:00:40.780
more things done if..."
And that's the important part, "If

00:00:41.120 --> 00:00:45.089
you have some experience as a developer,
if you can combine your knowledge

00:00:45.200 --> 00:00:49.140
with AI." And I've said that in many v-
videos and episodes and

00:00:49.260 --> 00:00:52.800
I don't wanna repeat all of that here,
but there's one thing

00:00:52.880 --> 00:00:56.800
that's missing maybe or one thing
that I also find important

00:00:56.840 --> 00:00:59.330
when we look at AI from a developer's
perspective.

00:00:59.340 --> 00:01:03.260
And that is the fact that AI, even

00:01:03.280 --> 00:01:07.200
if we are an experienced developer,
enables us to build

00:01:07.260 --> 00:01:10.360
things we previously weren't able to
build.

00:01:10.460 --> 00:01:13.840
And also important, it can help us

00:01:14.020 --> 00:01:17.120
learn new things and expand our skill

00:01:17.180 --> 00:01:20.640
set. Now I will say there also is the

00:01:20.680 --> 00:01:22.860
danger of going the opposite direction.

00:01:22.900 --> 00:01:26.560
There is the danger,
and I've talked about this too, of

00:01:27.120 --> 00:01:31.100
relying on AI too much,
of not understanding the code it produces,

00:01:31.160 --> 00:01:35.020
maybe of not really looking at it
and of ending up in a code base

00:01:35.060 --> 00:01:38.450
where you don't know the code,
where you don't understand what's going

00:01:38.520 --> 00:01:41.920
on. Uh,
that's definitely not the direction you

00:01:41.940 --> 00:01:45.700
But let's say I wanna build my own
web-based

00:01:45.920 --> 00:01:49.760
image editor,
something trivial where I maybe can just

00:01:49.820 --> 00:01:53.540
the saturation of an image and save
that as a new

00:01:53.580 --> 00:01:57.200
image.
Or I wanna build my own very basic video

00:01:57.240 --> 00:02:01.220
tool that, uh, encodes a video,
maybe extracts

00:02:01.260 --> 00:02:05.100
the audio from a video.
I wanna build my own tool because I wanna

00:02:05.140 --> 00:02:07.560
an overall workflow or anything like that.

00:02:07.600 --> 00:02:11.450
Now previously, I'll be honest, these
are things that would've

00:02:11.480 --> 00:02:15.320
taken me quite some time to build because
I had basically no

00:02:15.360 --> 00:02:19.070
experience with working with the Canvas
API or

00:02:19.260 --> 00:02:21.620
of image manipulation in the browser.

00:02:21.680 --> 00:02:25.520
I had no experience with MPEG.
So these are things where I

00:02:25.540 --> 00:02:29.380
would have to dive into the documentation,
read tutorials, watch videos.

00:02:29.400 --> 00:02:33.200
And whilst that all is fun
and builds up some knowledge and

00:02:33.260 --> 00:02:37.120
is therefore in theory important,
it's also

00:02:37.160 --> 00:02:41.100
something where I probably wouldn't have
had the time so I wouldn't have

00:02:41.120 --> 00:02:44.360
done it and I would've found other
solutions to my problem, so

00:02:45.080 --> 00:02:47.960
everything's good to some extent.

00:02:48.020 --> 00:02:51.840
But I do believe that it's now actually a

00:02:52.040 --> 00:02:55.840
better time to be a developer because we
can

00:02:55.940 --> 00:02:59.840
now build any kind of utility tool,
any kind

00:02:59.880 --> 00:03:03.100
of workflow we wanna build, even
if it involves

00:03:03.480 --> 00:03:06.660
things or skills that we don't have

00:03:06.820 --> 00:03:10.460
yet. Because if I need a, a tool that

00:03:10.500 --> 00:03:14.460
extracts audio from a video,
I can just prompt AI

00:03:14.560 --> 00:03:18.400
to build it for me
and I very likely get a tool

00:03:18.440 --> 00:03:22.180
that works. Is it perfect? No.
It will probably not cover all

00:03:22.300 --> 00:03:26.280
possible edge cases,
it will probably be buggy to

00:03:26.320 --> 00:03:29.150
some extent,
but it might be good enough for me.

00:03:29.180 --> 00:03:32.580
It might be good enough for me to get the
job done which I want to get done.

00:03:32.660 --> 00:03:36.320
If it fails on a four-hour video
but works on a ten-minute

00:03:36.400 --> 00:03:40.329
video and I only have ten-minute videos,
then this is

00:03:40.440 --> 00:03:43.340
not a bug or a problem I need to tackle.

00:03:43.420 --> 00:03:47.340
So I don't need the perfect tool,
I just need a tool that's good

00:03:47.360 --> 00:03:49.800
enough. But isn't that vibe coding?

00:03:49.820 --> 00:03:53.100
Well, to some degree it is,
but here's the interesting thing,

00:03:53.120 --> 00:03:56.860
me.
I really enjoy the process of building

00:03:56.920 --> 00:04:00.370
utility tools, little helpers that I can

00:04:00.420 --> 00:04:04.360
use and then I enjoy the part of diving
into the code

00:04:04.700 --> 00:04:07.020
and doing some research of what's
happening there.

00:04:07.040 --> 00:04:10.920
But since I already got something working,
since I already have

00:04:10.960 --> 00:04:14.860
that initial solution, I don't have

00:04:14.940 --> 00:04:18.640
to dive into the docs
and try to find out which parts are

00:04:18.660 --> 00:04:22.520
important for me. Instead,
I can just read more about

00:04:22.580 --> 00:04:26.420
the, the methods, the tools, the,
the functions that are

00:04:26.440 --> 00:04:29.750
being used in my code
and understand what exactly

00:04:29.760 --> 00:04:33.080
And that allows me to build up an
understanding of how that

00:04:33.090 --> 00:04:36.580
technology, that library, or whatever
is getting used in my code

00:04:37.100 --> 00:04:40.940
works under the hood.
And I've learned a ton about working with

00:04:41.000 --> 00:04:44.490
Canvas API, about web animations,
about tools like

00:04:44.540 --> 00:04:48.510
MPEG,
and many other things just because I built

00:04:48.580 --> 00:04:52.540
own little helpers. And these
are not just fun projects, by the way.

00:04:52.550 --> 00:04:56.480
These are actual things
that then in the end might find their way

00:04:56.540 --> 00:04:59.180
into an actual product I'm building.

00:04:59.260 --> 00:05:02.540
Obviously then polished and with, um,

00:05:03.360 --> 00:05:07.300
with more knowledge from my side then
because of course once I build something

00:05:07.340 --> 00:05:11.080
production for other people to use,
I 100% want to make

00:05:11.140 --> 00:05:14.660
sure that there are no bugs in there,
or at least no

00:05:14.820 --> 00:05:18.680
bugs, uh, I, I know about
and I wanna make sure that it's stable,

00:05:18.740 --> 00:05:21.180
that it's fast,
that I use common best practices.

00:05:21.200 --> 00:05:25.170
So of course I will then take
that time to polish it and

00:05:25.240 --> 00:05:28.860
to learn more about it.
But for getting started with the learning

00:05:28.940 --> 00:05:32.860
process or if we're doing some one-off
tasks, I think we're

00:05:32.900 --> 00:05:36.670
living in the best time possible because
AI is a big

00:05:36.700 --> 00:05:40.030
enabler there. By the way,
also for example, if we're talking

00:05:40.120 --> 00:05:44.010
about technologies
or programming languages we're not

00:05:44.040 --> 00:05:47.790
that good in. Now, for example,
I'm way better in

00:05:47.840 --> 00:05:51.590
JavaScript than I am in Python also
because I

00:05:51.620 --> 00:05:55.520
haven't written a lot of Python code for
the last three to four years.

00:05:55.660 --> 00:05:58.940
Lots of small utility scripts
and workflows,

00:05:58.980 --> 00:06:01.150
definitely, but nothing complex.

00:06:01.200 --> 00:06:05.000
So I don't have any expert or deep expert

00:06:05.080 --> 00:06:08.450
level knowledge there.
Something I wanna change again,

00:06:08.460 --> 00:06:12.180
But it is what it is. Here of course,
AI can also

00:06:12.320 --> 00:06:15.940
help me get stuff done and then either I

00:06:16.020 --> 00:06:19.900
remember the syntax I see
and I remember what's happening there or I

00:06:19.940 --> 00:06:23.600
can then again dive deeper.
And I think that's a

00:06:23.660 --> 00:06:27.480
very valuable thing
or advantage we get out of

00:06:27.540 --> 00:06:31.360
AI, because it allows us to do stuff
that otherwise would've been

00:06:31.420 --> 00:06:34.140
impossible or maybe just not worth our
time

00:06:34.640 --> 00:06:38.340
and also grow our skill base. I

00:06:38.360 --> 00:06:42.260
think that is the,
the important part

00:06:42.420 --> 00:06:45.730
pure vibe coding where you don't even look
at the code, where you don't care about

00:06:45.740 --> 00:06:49.550
the code. I care. I like learning.
And as I've said in other

00:06:49.640 --> 00:06:52.820
episodes as well,
I can only recommend going

00:06:52.840 --> 00:06:56.440
deep, uh, as a developer.
You want the deep

00:06:56.560 --> 00:07:00.360
knowledge, you wanna be better than AI,
you wanna be an

00:07:00.420 --> 00:07:04.280
expert in certain matters
and you wanna understand what AI is

00:07:04.300 --> 00:07:08.130
doing and always get to a level where you
can correct AI or

00:07:08.400 --> 00:07:12.120
steer AI.
Because as long as that's the case,

00:07:12.180 --> 00:07:15.960
you'll have, uh,
a reason to get employed or

00:07:16.000 --> 00:07:19.780
whatever your goals are.
You don't wanna be on the same level as

00:07:19.820 --> 00:07:22.360
You wanna be above it at least in certain
areas.

00:07:22.480 --> 00:07:26.280
Obviously with the broad knowledge AI has,
you can't beat it

00:07:26.360 --> 00:07:29.920
everywhere but you can beat it in certain
domains and it can be a

00:07:29.940 --> 00:07:32.200
great enabler.