Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
8:00
any of it works or what the terms are for the host
8:02
or anything. The boosts are
8:04
the exact opposite. So we gain a massive efficiency
8:07
from that support because the payments
8:09
are final immediately. The splits are in the RSS
8:11
feed. So they're transparent and documented for everybody. I've
8:13
had hosts claimed who got paid in cash that
8:15
they were never paid after they left. This
8:18
fixes that problem for us from a production side.
8:20
But because the payment is instant, there's
8:22
no like invoicing that needs to happen. And then
8:24
payment has to be received and deposited and then
8:27
checks get written to people like none of that
8:29
has to happen because when you
8:31
boost, the split happens instantly. So
8:33
there's a massive efficiency gain, too. It
8:36
gives a small, tiny team the
8:38
ability to remain nimble and efficient while
8:41
they're getting direct support from the audience over
8:43
a peer-to-peer system that has no middleman. And
8:46
it creates the audience as the customer for
8:48
the show. We should really refer to
8:50
you as the customer because you're buying the product. We're
8:53
putting it out there for free, but we're asking, hey, if you like it,
8:56
maybe pay for it from time to time. Whatever value
8:58
you think it's worth. That's the value for value system. That's
9:00
why the boost works so great. But I understand
9:02
some of you want the autopilot mode and we
9:04
appreciate that more than you can know. You could
9:06
become part of the QA team by going to
9:09
coder.show slash membership. You sign up to get an ad free
9:11
version of the show and the coderly, although I
9:13
haven't released it yet because it's kind of a downer. So
9:15
I've kind of been thinking maybe I shouldn't. But,
9:17
you know, assuming I don't hold
9:19
it back, you get the coderly. I guess I
9:21
guess I retain editorial oversight. Is that is
9:24
that fair? I consider myself the editor in chief. We
9:27
appreciate your support. However you
9:29
like to do it. Coder.show slash boost because, yep, you
9:32
can boost from the Web now or get a new podcast
9:34
app and boost directly from your client or
9:36
coder.show slash membership and set it on autopilot
9:39
and get the ad free version. So you don't have to hear this
9:41
anymore. Thank you so much for supporting
9:43
the show. However you do it. All
9:47
right. Now I saw this article you put in the doc and
9:49
I thought, I don't know if I completely
9:51
agree with this guy's take, but he kind of does
9:53
a takedown of why you shouldn't use A.I. to write
9:55
your test. In most cases, there's a couple exceptions and
9:58
I'm curious what it made you think. What drew your
10:00
attention? So it's funny. I don't agree
10:03
with his ultimate take but I agree with just about
10:05
every premise he used to get there. Okay,
10:07
all right So the
10:09
thing that often you know, it comes
10:12
out and he talks talking about testing very
10:14
generally Then he
10:16
goes into TDD which he has a fair point
10:19
It can often be a waste right if especially
10:21
if you're TDD zealot TDD is test driven development
10:24
I don't hear much about that anymore But you remember
10:26
a couple years ago Chris heard all these people being
10:28
like gotta write your tests first banner maybe we go
10:30
to testing we go to spend half my time writing
10:32
tests and Many
10:34
of those tests were very shallow, right like
10:37
adding type checking to Ruby This was
10:39
kind of huge in the rails community, but spread it
10:41
elsewhere I
10:43
totally agree with this dude that TDD can often be
10:45
a waste Regression and integration testing
10:48
are important right regression testing is exactly
10:50
what it sounds like I
10:52
changed something did I break features that have already
10:54
been accepted by the customer Integration
10:57
testing is how does this new component
10:59
interact with other components component being used
11:01
extremely loosely there That
11:03
makes tons of sense, right? You're you're altering the
11:05
structure of the system. So you need
11:07
to make sure that that doesn't create any kind of side
11:10
effects, right now he had said yeah kind
11:13
of the philosophy was is that You
11:16
want to create tests that are testing the
11:18
intent of your code Not the actual function
11:20
of your code because otherwise
11:22
you end up just testing if the code changes or
11:24
not Do you think that so he's
11:26
saying this is what draw is drawing people to AI
11:29
is to create tests that test the intention
11:31
of the code Are
11:33
we seeing this? Where is this actually happening in the
11:35
real world? I actually haven't seen this at all I've
11:38
seen I've seen like, you know, co-pilot type
11:40
stuff where it's helping you write I've seen
11:43
Troubleshooting, but I haven't seen people that
11:45
are generating tests whole cloth using any
11:47
kind of AI generation Yeah,
11:50
I don't I mean co-pilot can definitely
11:53
Yeah, could generate a test. Yeah, if you
11:56
if you've written the method you want to
11:58
test first again, just like what anything else
12:00
co-pilot, you shouldn't just take
12:02
its suggestion wholesale and run. Right.
12:05
And he, I mean, I guess if you want to like,
12:08
make it a straw man argument, you could
12:10
say, well, that's what people are doing. I
12:13
don't know it if like, I'm curious to the audience,
12:15
is anybody doing this? Because
12:17
I don't think I know
12:19
anyone who is. I
12:21
could see areas where it would make
12:24
sense, right? You could use AI to
12:26
fuzz your app. You could use AI
12:28
to check translations potentially. So
12:31
you could see where you could generate usage
12:33
of an application maybe with AI a little
12:35
bit or check the language inside your app
12:37
automatically. I could see areas where you could
12:39
apply it, but none of it's quite
12:42
clear to me yet. I don't know of any practical demonstrations
12:45
other than like co-pilot generating something, you know,
12:47
in line. Yeah. And
12:49
I also like question, what is the difference between the intent of
12:51
the code and the function of the code? Because
12:55
I guess what he's saying is you
12:57
could write a test that pass because it
12:59
does what the code mechanically does, but it's
13:01
the wrong thing. Yeah. And
13:04
that's kind of where like project management and
13:06
just like knowing what features your customer or
13:08
your product actually want is. Yeah.
13:11
I will say that I have generated some tests
13:13
by writing out very like, explain it to
13:15
the AI, like it's five user
13:17
stories, right? Like the user should be able
13:19
to calculate the holiday hours or whatever. And
13:23
then what I do is I take that as a base
13:25
and I just make it a little, you know, I harden
13:28
it up, right? I make it a
13:30
little less shallow. But I'm not copy-possing
13:33
random AI-generated crap. And
13:35
I don't think anyone
13:37
is that
13:40
is not like super junior and just learning.
13:43
I did a little live research. So I don't
13:45
think it's gone much beyond that you can do
13:47
a thing. But on
13:49
February 14th of this year, Meta
13:53
submitted a study on automated unit
13:56
tests improvement using large language models
13:58
at Meta. That's
34:00
great. Thank you. That makes this, this is
34:02
probably our best earning episode of the year.
34:04
So thank you, episode. He says, hey guys,
34:06
I fell behind on my podcast after two
34:08
episodes of low support on the CODO program.
34:10
I was saddened. So here are
34:12
some boosts to boost you up. Oh,
34:15
you're supposed to use a bomb that
34:17
boost. That's great.
34:20
Thank you, sir. Really
34:23
appreciate that. Ty Alaskan comes in
34:25
with 31,337. I
34:28
hoard that with your kind covenant.
34:31
And you're going to appreciate this one, I think. I
34:33
think he makes a fair point and I have a
34:36
make good for him as long as he approves. Rust
34:39
may get the cool sound effects on JB,
34:41
but I'm boosting in to share the good
34:43
word about Golang. I work professionally
34:46
with a fair, with a fair few
34:48
other languages such as Python and Java
34:50
and JS, but I'm definitely
34:52
more productive in Go. All
34:55
right. He's winning me over with this argument. I
34:57
hereby decree with the full force of
35:00
all authority of an anonymous internet person,
35:02
that Golang is the official language of
35:04
the CODO radio program until somebody boosts
35:06
a higher amount in another language. Mexico.
35:08
Wow. 31,337 sats to beat that
35:14
until we have to live with Golang is the
35:16
official language. See, how do I boost? 31,338
35:21
objectives. Nailed it. Now,
35:23
Ty Alaskan, with your approval, we now have
35:26
an official Go
35:28
Boost. Mexico. So there you have it.
35:31
I'm surprised somebody else hasn't come up with that. Thank
35:33
you very much. Big fans of the
35:36
Go programming language here at the show. You know, that's
35:38
why we have the Go sound effects. Mexico. Because we're
35:40
big Go fans here. That's why we've always had that
35:42
forever. DG at PTC
35:44
comes in with a Jar Jar Boost, 5,000
35:46
sats. You're so boost.
35:49
I got burned by chat GPT at work
35:51
recently. Apple introduced bugs
35:53
in a recent macOS release that have been then
35:56
broke some Docker builds that I rely
35:58
on using Rosetta. has
40:00
the momentum. Yeah. So if
40:02
somebody somebody could link me to something
40:04
I could read or give me a
40:06
brief on how could it get this
40:08
dominant position. Who did they lay
40:11
down with. Was it unreal. Like who
40:13
was it. I don't. I mean I don't know. I
40:15
came along by the time I started paying attention.
40:17
Kuda was everywhere and it was a problem on
40:20
Linux. That's when I started paying attention. Yeah. Also
40:22
the equity fine thing would never ever happen. I'm
40:24
the lawyer but I'm pretty sure the Fourth Amendment
40:26
says no to that. Okay.
40:29
Well thank you everybody. We had nine boosters. Of
40:32
course we have the 2000 sat cut off for
40:34
time but we stacked a fantastic five hundred and
40:36
sixty eight thousand five hundred and fifty eight sats.
40:41
Nice. Quality. Thank
40:44
you everybody. You can now boost
40:46
from the web if you go to coder dot show
40:48
slash boost you just need something like strike which is
40:50
now available in over a hundred countries just launched in
40:52
the UK or something like the cash app. I really
40:54
like anything that just scans the lightning code
40:58
and then you can send a boost from the web or get
41:00
all the new podcasting to features like value
41:03
for value integrated in their
41:05
transcripts chapters live streams notification
41:08
of release within 90 seconds and a
41:11
whole lot more including better and
41:13
improved search discovery and directories new
41:16
podcast apps.com for all
41:19
of that. I wanted to give a little public service
41:21
announcement before we get out of here because you know
41:23
I care about you and your
41:25
entertainment. I know you
41:27
got some kids the Ruby people strange
41:29
people started using get suddenly just exploded.
41:31
Now how did he get in here.
41:34
Who let him in here. I
41:36
don't know who let him in here. Wait I could
41:38
just see I could just see Linus walking into like
41:41
a hipster coffee shop just in horror
41:43
what all the Ruby hipsters like 2004 like what the hell is
41:45
happening to me. I
41:50
wanted you and all the dads and anybody who's just
41:52
a Star Trek fan to step
41:54
up this week because Star Trek prodigy
41:57
season two dropped on Netflix and it's
41:59
Netflix styles
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More