WEBVTT

00:00:00.080 --> 00:00:03.740
At this year's World Economic Forum in
Davos, Dario

00:00:03.780 --> 00:00:07.640
Amodei predicted that within the next 12

00:00:07.720 --> 00:00:11.380
months or so, AI would be able to write

00:00:11.600 --> 00:00:15.400
all code, fully automated, on its own,
essentially.

00:00:15.410 --> 00:00:19.280
And it's worth paying attention to what
this man has to say and I'll

00:00:19.380 --> 00:00:22.920
also share my thoughts on it
and why I think that you should take a

00:00:23.060 --> 00:00:27.020
nuanced look at that.
Not just because Dario, of course,

00:00:27.060 --> 00:00:30.670
is the CEO of Anthropic which
is one of the most important players in

00:00:30.740 --> 00:00:34.540
generative AI space, especially
when it comes to models related to

00:00:34.560 --> 00:00:38.140
coding.
But it's also worth paying attention

00:00:38.220 --> 00:00:42.140
year, in the beginning of 2025,
he predicted

00:00:42.150 --> 00:00:45.880
that AI would be able to write 90% of all
code

00:00:45.920 --> 00:00:49.800
within the next three or six months
and depending on how

00:00:49.880 --> 00:00:52.700
you look at that, he
was not entirely wrong.

00:00:52.720 --> 00:00:56.400
Now, sure, this timeline is probably,
or was

00:00:56.500 --> 00:01:00.420
probably a bit too aggressive.
It maybe wasn't three or

00:01:00.580 --> 00:01:04.540
six months, more like six to nine months,
not sure.

00:01:04.599 --> 00:01:08.500
And, of course,
it totally depends on your field, so in

00:01:08.540 --> 00:01:12.130
which area you're programming,
also in the programming language you're

00:01:12.160 --> 00:01:16.020
It will depend on your company,
the policies you have in your company.

00:01:16.040 --> 00:01:19.260
And of course, it also depends on, on you,
your

00:01:19.380 --> 00:01:22.500
preferences, your experiences with AI.

00:01:22.640 --> 00:01:26.200
I can say that for me,
I am at a point I would say since

00:01:26.480 --> 00:01:30.440
maybe December, November, where for some

00:01:30.480 --> 00:01:33.660
projects, most of my projects,
AI does write

00:01:34.260 --> 00:01:38.040
80, 90% of all code for me probably.

00:01:38.060 --> 00:01:42.040
But this is now where things get
interesting.

00:01:42.200 --> 00:01:45.930
I'm not talking about wipe coding
and I'm not

00:01:45.960 --> 00:01:49.610
talking about AI doing it all on its own
and I'm also not

00:01:49.640 --> 00:01:53.240
talking about AI getting it all right.

00:01:53.300 --> 00:01:56.400
Instead,
things are way more difficult than that.

00:01:56.420 --> 00:02:00.200
So it does absolutely not do that

00:02:00.320 --> 00:02:03.850
on its own. It also is

00:02:04.000 --> 00:02:07.890
not about wipe coding here. I mean,

00:02:07.920 --> 00:02:11.840
you can probably wipe code
but I've shared my view on that before

00:02:11.880 --> 00:02:15.500
me, by the way,
just to make sure we all got the same

00:02:15.600 --> 00:02:18.640
really means that you don't look at the
code at all, that you don't care about the

00:02:18.700 --> 00:02:21.660
code,
that you don't necessarily even know how

00:02:21.700 --> 00:02:25.620
So yeah, this is also not about wipe

00:02:25.720 --> 00:02:29.630
coding here. Instead, I would say we are

00:02:29.680 --> 00:02:33.580
at a point where you can hand off many
tasks

00:02:33.660 --> 00:02:37.560
to AI with a good plan. So

00:02:37.620 --> 00:02:40.460
if you have a good plan, AI can

00:02:40.680 --> 00:02:44.600
implement it. So I think we're at that

00:02:44.720 --> 00:02:48.680
point right now. You,
you can definitely do that, at least

00:02:48.690 --> 00:02:52.660
for some tasks. But of course, you wanna

00:02:52.720 --> 00:02:56.580
review that code and, uh, I know there are

00:02:56.600 --> 00:03:00.320
people that tell you you don't need to do
that anymore. Might work for them.

00:03:00.340 --> 00:03:04.220
Definitely not working for me and
if you're shipping anything, if you're

00:03:04.360 --> 00:03:07.820
shipping anything to customers,
you're responsible for that.

00:03:07.960 --> 00:03:11.920
You, as a developer, you can't say, "Oh,
the AI got that wrong." No, it's

00:03:12.040 --> 00:03:15.920
your responsibility
and I won't take responsibility in code

00:03:16.580 --> 00:03:19.980
I haven't reviewed, I don't understand.

00:03:20.080 --> 00:03:23.320
Also, still, AI makes many

00:03:23.380 --> 00:03:27.320
mistakes, so I need to fix mistakes

00:03:27.780 --> 00:03:30.640
or steer AI in the right

00:03:31.500 --> 00:03:32.340
direction.

00:03:34.620 --> 00:03:38.200
And that is also very important. That's,
of course,

00:03:38.240 --> 00:03:42.000
important to the planning part
or related to the

00:03:42.040 --> 00:03:44.580
planning part,
that you steer AI in the right direction.

00:03:44.600 --> 00:03:48.030
But it's also related to what you do with
code AI gives you.

00:03:48.060 --> 00:03:51.760
So very far away from AI

00:03:52.140 --> 00:03:56.040
generating 90% of the code and
that means I have

00:03:56.080 --> 00:04:00.040
no work to do. Far away from that.
It just means that I

00:04:00.080 --> 00:04:03.560
got a very fast typer who can

00:04:03.620 --> 00:04:07.460
implement my plans,
but those plans need to be good

00:04:07.960 --> 00:04:11.240
and the output is kind of varying in
quality.

00:04:11.320 --> 00:04:15.240
Still, for me,
it's very likely faster than

00:04:15.280 --> 00:04:18.920
it all from scratch.
If I instead try to build good plans,

00:04:19.000 --> 00:04:22.980
split that up into smaller chunks,
let AI write the code for me and

00:04:23.020 --> 00:04:24.800
then review and fine-tune that code.

00:04:24.840 --> 00:04:28.430
And by the way, with planning,
I mean really detailed

00:04:28.430 --> 00:04:32.360
plans where I also break down the exact
libraries I wanna use, the

00:04:32.420 --> 00:04:35.680
patterns I wanna use,
the architecture of the software I wanna

00:04:35.760 --> 00:04:39.740
step-by-step. So it's not a,
a rough plan or a general plan; it's

00:04:39.780 --> 00:04:43.500
a very detailed plan.
That works for me and therefore, for

00:04:43.560 --> 00:04:47.380
me, with, uh,
with these restrictions here in mind, I

00:04:47.420 --> 00:04:50.400
would say, yeah, it can probably write

00:04:50.660 --> 00:04:54.469
90%-ish of the code. But that does not

00:04:54.500 --> 00:04:56.760
mean that it does 90% of the work on its
own.

00:04:56.780 --> 00:05:00.550
Now,
that's important context because that's

00:05:00.600 --> 00:05:03.890
to last year's statement by Dario.

00:05:03.920 --> 00:05:07.820
So yeah, I would say we are kind of there,
but not in the sense of AI doing

00:05:07.860 --> 00:05:11.380
it all on its own. Now,
what about this year's

00:05:11.480 --> 00:05:15.420
statement? Though, in this year,
Dario essentially said that AI will be

00:05:15.480 --> 00:05:19.460
able to do what software engineers do
and write software fully on its own,

00:05:19.480 --> 00:05:22.360
end-to-end, within six to 12 months.

00:05:22.400 --> 00:05:25.920
That's this year's statement. Now,
I fully recommend

00:05:25.960 --> 00:05:29.920
watching this entire talk.
It's very interesting, but I have some

00:05:29.940 --> 00:05:33.850
thoughts about this statement.
And obviously, just to make

00:05:33.860 --> 00:05:37.460
and obvious, I'm in no way smarter
or more capable

00:05:37.940 --> 00:05:41.280
in judging the performance of AI models
than Dario.

00:05:41.320 --> 00:05:45.020
But I'm also not the CEO of a company
that needs to

00:05:45.120 --> 00:05:48.710
sell these AI models
and I can share from my own

00:05:48.740 --> 00:05:52.310
experiences. Now, as I said,
I would agree with last year's

00:05:52.360 --> 00:05:56.020
statement to some degree,
but with many caveats, with many

00:05:56.300 --> 00:06:00.240
restrictions. The AI
is definitely not writing 90% of my code

00:06:00.280 --> 00:06:04.040
its own. So naturally,
I have a very hard time

00:06:04.100 --> 00:06:07.976
imagining... that this will come true,
and I have a very hard time

00:06:08.016 --> 00:06:11.916
imagining that this will come true,
not just within the next 6 or 12

00:06:11.996 --> 00:06:15.616
months, but anywhere in the near future.

00:06:15.626 --> 00:06:19.016
Now, I totally see that AI is

00:06:19.036 --> 00:06:22.736
capable of building software on its

00:06:22.856 --> 00:06:26.046
own in a loop,
with the Rolf loop that's getting a lot of

00:06:26.076 --> 00:06:29.956
hype, uh, related to Claude code.
I totally see

00:06:30.016 --> 00:06:33.916
that. But the full work of a software

00:06:33.956 --> 00:06:36.916
engineer includes the tasks I outlined
here

00:06:37.576 --> 00:06:41.396
and arguably more than that.
It includes building a

00:06:41.456 --> 00:06:44.776
good plan, defining which architecture,
which

00:06:44.916 --> 00:06:48.396
patterns, which technologies to use,
reviewing the

00:06:48.456 --> 00:06:52.256
code, and also, of course, analyzing the

00:06:52.296 --> 00:06:56.216
code, fixing problems with the code,
taking responsibility for that

00:06:56.296 --> 00:07:00.036
code.
That is something I have a very hard time

00:07:00.636 --> 00:07:04.436
within the near future because right now,
the AI I can

00:07:04.556 --> 00:07:07.776
use is too far away from that. It's a

00:07:07.876 --> 00:07:11.836
talented, fast writer
that makes a lot of mistakes along

00:07:11.856 --> 00:07:14.816
the way and that needs very clear
guidance.

00:07:14.836 --> 00:07:18.636
And that going to models
that can do it all on its own, that can

00:07:18.676 --> 00:07:22.336
plan out entire architectures cleanly on
their

00:07:22.416 --> 00:07:25.666
own, that are capable of using the latest

00:07:25.856 --> 00:07:29.756
technologies,
of writing error-free secure code totally

00:07:29.816 --> 00:07:33.336
on their own without review
or only with review by other

00:07:33.396 --> 00:07:37.176
AIs,
that is something I have a very hard time

00:07:37.516 --> 00:07:41.476
seeing when I look at today's models
and also when I look at the

00:07:41.536 --> 00:07:44.106
progress of models over the last years.

00:07:44.156 --> 00:07:47.676
Because, yeah, that progress, of course,
has been remarkable.

00:07:47.716 --> 00:07:51.556
It has been steady and, and good,
and especially the tools I shared that

00:07:51.596 --> 00:07:55.576
before got way better.
So I shared it before

00:07:55.656 --> 00:07:59.356
with models. I'm not sure
if we're the raw model intelligence,

00:07:59.416 --> 00:08:02.806
we're still on a linear,
leave alone exponential

00:08:02.856 --> 00:08:06.836
trajectory. For the tools,
I would say we definitely are on some

00:08:06.896 --> 00:08:10.876
linear trajectory. But I have a hard

00:08:10.956 --> 00:08:14.745
time seeing that being enough to get us to
that full

00:08:14.796 --> 00:08:18.266
automation anytime soon. Now,

00:08:18.316 --> 00:08:22.076
naturally, these CEOs have other interests

00:08:22.136 --> 00:08:26.025
than,
than me and you could say I have the

00:08:26.076 --> 00:08:29.636
us developers also because I also sell
programming

00:08:29.676 --> 00:08:33.656
courses. But let's be totally honest,
that is not my role

00:08:33.696 --> 00:08:37.666
and nothing I will be able to do.
I'm just sharing here what

00:08:37.716 --> 00:08:41.525
my experience with AI has been,
and I'm very open-minded, I would

00:08:41.536 --> 00:08:43.805
say, for AI. I'm using it a lot,
as I said.

00:08:43.836 --> 00:08:47.546
It does write 90% of my code.
It's just far away from

00:08:47.596 --> 00:08:51.496
that full automation. But yeah,
please let me know your

00:08:51.516 --> 00:08:55.376
thoughts on this as well,
your experiences with AI,

00:08:55.436 --> 00:08:58.616
working because you might not be using AI
at all or only for

00:08:59.176 --> 00:09:03.096
small isolated tasks
and you might still write most of the code

00:09:03.316 --> 00:09:07.146
or you're using it for everything,
you're not even looking on the c- uh,

00:09:07.196 --> 00:09:09.536
code, and you're having great results.

00:09:09.656 --> 00:09:12.805
I'm interested to learn more about that,
so please let me know.

00:09:12.856 --> 00:09:15.055
And yeah, have a great time