РАСШИФРОВКА 🔗
WEBVTT
00:00:00.020 --> 00:00:03.540
Python in 2025 is a delightfully refreshing place.
00:00:04.080 --> 00:00:05.100
The guild's days are numbered,
00:00:05.560 --> 00:00:07.220
packaging is getting sharper tools,
00:00:07.620 --> 00:00:09.200
and the type checkers are multiplying
00:00:09.400 --> 00:00:11.040
like gremlins snacking after midnight.
00:00:11.500 --> 00:00:13.720
On this episode, we have an amazing panel
00:00:13.920 --> 00:00:15.500
to give us a range of perspectives
00:00:15.940 --> 00:00:18.480
on what mattered in 2025 in Python.
00:00:19.040 --> 00:00:20.940
We have Barry Warsaw, Brett Cannon,
00:00:21.380 --> 00:00:24.160
Gregory Kapfhammer, Jody Burchell, Reuven Lerner,
00:00:24.580 --> 00:00:27.480
and Thomas Wouters on the show to give us their thoughts.
00:00:28.040 --> 00:00:33.260
This is Talk Python To Me, episode 532, recorded December 9th, 2025.
00:00:50.460 --> 00:00:55.180
Welcome to Talk Python To Me, the number one Python podcast for developers and data scientists.
00:00:55.700 --> 00:00:57.100
This is your host, Michael Kennedy.
00:00:57.480 --> 00:01:01.020
I'm a PSF fellow who's been coding for over 25 years.
00:01:01.660 --> 00:01:02.780
Let's connect on social media.
00:01:03.100 --> 00:01:06.180
You'll find me and Talk Python on Mastodon, Bluesky, and X.
00:01:06.540 --> 00:01:08.360
The social links are all in your show notes.
00:01:09.140 --> 00:01:12.640
You can find over 10 years of past episodes at talkpython.fm.
00:01:12.820 --> 00:01:16.020
And if you want to be part of the show, you can join our recording live streams.
00:01:16.380 --> 00:01:16.860
That's right.
00:01:17.160 --> 00:01:20.300
We live stream the raw uncut version of each episode on YouTube.
00:01:20.660 --> 00:01:25.320
Just visit talkpython.fm/youtube to see the schedule of upcoming events.
00:01:25.530 --> 00:01:29.200
Be sure to subscribe there and press the bell so you'll get notified anytime we're recording.
00:01:29.940 --> 00:01:32.740
Look into the future and see bugs before they make it to production.
00:01:33.480 --> 00:01:38.840
Sentry's SEER AI code review uses historical error and performance information at Sentry
00:01:39.150 --> 00:01:43.160
to find and flag bugs in your PRs before you even start to review them.
00:01:43.840 --> 00:01:45.660
Stop bugs before they enter your code base.
00:01:46.160 --> 00:01:50.040
Get started at talkpython.fm/seer-code-review.
00:01:50.880 --> 00:01:58.920
Hey, before we jump into the interview, I just want to send a little message to all the companies out there with products and services trying to reach developers.
00:01:59.620 --> 00:02:01.180
That is the listeners of this show.
00:02:01.680 --> 00:02:05.500
As we're rolling into 2026, I have a bunch of spots open.
00:02:05.600 --> 00:02:12.360
So please reach out to me if you're looking to sponsor a podcast or just generally sponsor things in the community.
00:02:12.960 --> 00:02:14.500
And you haven't necessarily considered podcasts.
00:02:14.700 --> 00:02:19.620
You really should reach out to me and I'll help you connect with the Talk Python audience.
00:02:20.600 --> 00:02:27.280
thanks everyone for listening all of 2025 and here we go into 2026 cheers hey everyone it's so
00:02:27.370 --> 00:02:31.560
awesome to be here with you all thanks for taking the time out of your day to be part of talk python
00:02:31.820 --> 00:02:37.740
for this year in review this python year in review so yeah let's just jump right into it gregory
00:02:38.180 --> 00:02:42.500
welcome welcome to the show welcome back to the show how you doing hi i'm an associate professor
00:02:42.660 --> 00:02:46.559
of computer and information science and i do research and software engineering and software
00:02:46.580 --> 00:02:51.720
testing. I've built a bunch of Python tools, and one of the areas we're studying now is flaky test
00:02:51.920 --> 00:02:57.720
cases in Python projects. I'm also really excited about teaching in a wide variety of areas. In fact,
00:02:57.720 --> 00:03:02.960
I use Python for operating systems classes or theory of computation classes. And one of the
00:03:02.960 --> 00:03:08.560
things I'm excited about is being a podcast host. I'm also a host on the Software Engineering Radio
00:03:08.820 --> 00:03:14.540
podcast sponsored by the IEEE Computer Society, and I've had the cool opportunity to interview a
00:03:14.540 --> 00:03:18.460
whole bunch of people in the Python community. So Michael, thanks for welcoming me to the show.
00:03:18.640 --> 00:03:22.640
Yeah, it's awesome to have you back. And we talked about FlakyTest last time. I do have to say
00:03:23.100 --> 00:03:29.140
your AV setup is quite good. I love the new mic and all that. Thomas, welcome. Awesome to have
00:03:29.160 --> 00:03:34.100
you here. Thanks for having me. I'm Thomas Wouters. I'm a longtime Python core developer,
00:03:34.300 --> 00:03:40.320
although not as long as one of the other guests on this podcast. I worked at Google for 17 years.
00:03:40.460 --> 00:03:45.380
for the last year or so I've worked at Meta. In both cases, I work on Python itself within the
00:03:45.600 --> 00:03:50.920
company and just deploying it internally. I've also been a board member of the PSF, although I'm not
00:03:51.100 --> 00:03:58.020
one right now. And I've been a steering council member for five years and currently not because
00:03:58.280 --> 00:04:02.380
the elections are going and I don't know what the result is going to be. But I think there's like
00:04:02.820 --> 00:04:09.099
five, six chance that I'll be on the steering council since we only have six candidates for
00:04:09.120 --> 00:04:14.680
five positions when this episode probably airs. I don't know. That's quite the contribution to the
00:04:14.740 --> 00:04:19.299
whole community. Thank you. I always forget this. I also got the, what is it, the Distinguished
00:04:19.400 --> 00:04:23.780
Service Award from the PSF this year. I should probably
Load full transcript
mention that. So yes, I have been
00:04:24.040 --> 00:04:29.680
recognized. No need to talk about it further. Wonderful. Wonderful. Jody, welcome back on the
00:04:29.840 --> 00:04:34.039
show. Awesome to catch up with you. Yeah, thanks for having me back. I am a data scientist and
00:04:34.060 --> 00:04:39.560
developer advocate at JetBrains working on PyCharm. And I've been a data scientist for around 10
00:04:39.840 --> 00:04:45.220
years. And prior to that, I was actually a clinical psychologist. So that was my training,
00:04:45.610 --> 00:04:49.860
my PhD, but abandoned academia for greener pastures. Let's put it that way.
00:04:50.190 --> 00:04:50.860
Noah Franz-Gurig.
00:04:53.180 --> 00:04:55.000
Brett, hello. Good to see you.
00:04:55.100 --> 00:05:00.200
Hello. Yes. Yeah, let's see here. I've been at Microsoft for 10 years. I started doing,
00:05:00.300 --> 00:05:03.080
working on AI R&D for Python developers.
00:05:03.880 --> 00:05:06.040
Also keep Wazzy running for Python here
00:05:06.380 --> 00:05:09.140
and do a lot of internal consulting for teams outside.
00:05:09.860 --> 00:05:11.880
I am actually the shortest running
00:05:12.700 --> 00:05:14.060
core developer on this call, amazingly,
00:05:14.200 --> 00:05:15.660
even though I've been doing it for 22 years.
00:05:16.060 --> 00:05:17.840
I've also only gotten the Frank Willison Award,
00:05:18.080 --> 00:05:18.820
not the DSA.
00:05:19.300 --> 00:05:20.900
So I feel very under accomplished here
00:05:20.900 --> 00:05:21.600
as a core developer.
00:05:22.040 --> 00:05:23.700
Yeah, that's me in a nutshell.
00:05:24.060 --> 00:05:25.900
Otherwise, I'm still trying to catch that.
00:05:26.940 --> 00:05:27.460
Most quoted.
00:05:27.900 --> 00:05:29.160
Yeah, most quoted.
00:05:29.220 --> 00:05:33.360
I will say actually at work, it is in my email footer that I'm a famous Python quotationist.
00:05:33.700 --> 00:05:35.660
That was Anthony Shaw's suggestion, by the way.
00:05:35.860 --> 00:05:39.600
That was not mine, but does link to the April Fool's joke from last year.
00:05:39.850 --> 00:05:42.920
And I am still trying to catch Anthony Shaw, I think, on appearances on this podcast.
00:05:43.060 --> 00:05:44.120
Well, plus one.
00:05:44.850 --> 00:05:46.060
Anthony Shaw should be here, honestly.
00:05:46.440 --> 00:05:47.540
I mean, I put it out into Discord.
00:05:48.520 --> 00:05:50.740
He could have been here, but probably at an odd time.
00:05:51.000 --> 00:05:54.820
You used to work on VS Code a bunch on the Python aspect of VS Code.
00:05:54.870 --> 00:05:56.800
You recently changed roles, right?
00:05:57.000 --> 00:05:57.520
Not recently.
00:05:57.560 --> 00:05:59.500
That was, I used to be the dev manager.
00:05:59.500 --> 00:06:00.880
Every seven years, years ago.
00:06:01.100 --> 00:06:02.580
Yeah, September of 2024.
00:06:03.530 --> 00:06:04.200
So it's been over a year.
00:06:04.330 --> 00:06:04.940
But yeah, I used to be the dev manager.
00:06:04.940 --> 00:06:06.040
That counts as recent for me.
00:06:06.240 --> 00:06:09.400
Yes, I used to be the dev manager for the Python experience in VS Code.
00:06:09.520 --> 00:06:10.340
Okay, very cool.
00:06:10.540 --> 00:06:11.340
That's quite a shift.
00:06:11.480 --> 00:06:12.820
Yeah, it went back to being an IC basically.
00:06:13.820 --> 00:06:16.220
You got all, you're good at your TPS reports again now?
00:06:17.500 --> 00:06:19.480
Actually, I just did do my connect, so I kind of did.
00:06:19.580 --> 00:06:19.900
Awesome.
00:06:20.800 --> 00:06:23.480
Reuven, I bet you haven't filed a TPS report in at least a year.
00:06:23.660 --> 00:06:24.420
So yeah, I'm Reuven.
00:06:24.600 --> 00:06:27.960
I'm a freelance Python and Pandas trainer.
00:06:28.460 --> 00:06:31.860
I just celebrated this past week 30 years since going freelance.
00:06:32.660 --> 00:06:34.340
So I guess it's working out okay.
00:06:35.700 --> 00:06:37.720
We'll know at some point if I need to get a real job.
00:06:38.020 --> 00:06:41.740
I teach Python Pandas both at companies and on my online platform.
00:06:41.970 --> 00:06:42.660
I have newsletters.
00:06:42.850 --> 00:06:47.160
I've written books, speaking conferences, and generally try to help people improve their
00:06:47.590 --> 00:06:51.660
Python and Pandas fluency and confidence and have a lot of fun with this community as well
00:06:51.660 --> 00:06:52.200
as with the language.
00:06:52.520 --> 00:06:53.260
Oh, good to have you here.
00:06:53.480 --> 00:06:55.840
Barry, it's great to have a musician on the show.
00:06:56.840 --> 00:06:57.200
Thanks.
00:06:58.340 --> 00:06:59.780
Yeah, I've got my bases over here.
00:07:00.060 --> 00:07:02.200
So, you know, if you need to be serenaded.
00:07:02.740 --> 00:07:05.280
Yeah, like a Zen of Python may break out at any moment.
00:07:05.380 --> 00:07:06.640
You never know when it's going to happen.
00:07:07.000 --> 00:07:07.880
Thanks for having me here.
00:07:08.160 --> 00:07:12.140
Yeah, I've been a core developer for a long time, since 1994.
00:07:13.840 --> 00:07:19.240
And I've been, you know, in the early days, I did tons of stuff for Python.org.
00:07:20.000 --> 00:07:27.060
I worked with Guido at CNRI and we moved everything from the mailing, the Postmaster stuff,
00:07:27.580 --> 00:07:32.480
and the version control systems back in the day, websites, all that kind of stuff.
00:07:32.560 --> 00:07:34.720
I try to not do any of those things anymore.
00:07:35.500 --> 00:07:38.500
There's way more competent people doing that stuff now.
00:07:39.120 --> 00:07:41.120
I have been a release manager.
00:07:42.920 --> 00:07:46.860
I'm currently back on the steering council and running again.
00:07:47.940 --> 00:07:52.220
between Thomas and I, we'll see who makes it to six years, I guess.
00:07:52.880 --> 00:07:55.460
And I'm currently working for NVIDIA,
00:07:56.050 --> 00:07:57.780
and I do all Python stuff.
00:07:58.680 --> 00:08:01.860
Some half and half, roughly, of internal things
00:08:02.120 --> 00:08:05.140
and external open source community work,
00:08:05.600 --> 00:08:08.160
both in packaging and in core Python.
00:08:08.700 --> 00:08:10.800
That's, I guess, I think that's about it.
00:08:10.920 --> 00:08:14.660
Yeah, you all are living in exciting tech spaces, that's for sure.
00:08:14.900 --> 00:08:15.520
That's for sure.
00:08:15.520 --> 00:08:15.960
For sure.
00:08:16.160 --> 00:08:20.780
Yeah. Well, great to have you all back on the show. Let's start with our first topic. So the
00:08:20.850 --> 00:08:26.740
idea is we've each picked at least a thing that we think stood out in 2025 in the Python space
00:08:27.580 --> 00:08:33.200
that we can focus on. And let's go with Jody first. I'm excited to hear what you thought was
00:08:33.360 --> 00:08:40.080
one of the bigger things. I'm going to mention AI. Like, wow, what a big surprise. So to kind of
00:08:40.140 --> 00:08:45.360
give context of where I'm coming from, I've been working in NLP for a long time. I like to say I was
00:08:45.260 --> 00:08:50.180
working on LLMs before they were cool. So sort of playing around with the very first releases from
00:08:50.320 --> 00:08:56.400
Google in like 2019, incorporating that into search. So I've been very interested sort of
00:08:56.620 --> 00:09:03.160
seeing the unfolding of the GPT models as they've grown. And let's say slightly disgusted by the
00:09:03.520 --> 00:09:09.519
discourse around the models as they become more mainstream, more sort of the talk about people's
00:09:09.540 --> 00:09:15.200
jobs being replaced, a lot of the hysteria, a lot of the doomsday stuff. So I've been doing talks
00:09:15.310 --> 00:09:19.820
and other content for around two and a half years now, just trying to cut through the hype a bit,
00:09:19.900 --> 00:09:23.660
being like, you know, they're just language models, they're good for language tasks. Let's think about
00:09:23.860 --> 00:09:29.360
realistically what they're about. And what was very interesting for me this year, I've been
00:09:29.620 --> 00:09:34.759
incorrectly predicting the bubble bursting for about two and a half years. So I was quite vindicated
00:09:34.840 --> 00:09:37.600
when in August, GPT-5 came out,
00:09:38.030 --> 00:09:40.420
and all of a sudden, everyone else started saying,
00:09:40.500 --> 00:09:41.500
maybe this is a bubble.
00:09:41.740 --> 00:09:43.800
Don't you think that was the first big release
00:09:44.060 --> 00:09:46.940
that was kind of a letdown compared to what the hype was?
00:09:47.100 --> 00:09:48.200
Yeah, and it was really interesting.
00:09:48.560 --> 00:09:50.980
So I found this really nice Atlantic article,
00:09:51.110 --> 00:09:52.540
and I didn't save it, unfortunately,
00:09:52.960 --> 00:09:56.140
but essentially it told sort of the whole story
00:09:56.350 --> 00:09:57.760
of what was going on behind the scenes.
00:09:58.000 --> 00:10:01.820
So GPT-4 came out in March of 2023,
00:10:02.390 --> 00:10:03.939
and that was
the model that came out
00:10:03.960 --> 00:10:08.520
with this Microsoft research paper saying, you know, sparks of AGI, artificial general intelligence,
00:10:08.660 --> 00:10:14.740
blah, blah, blah. And from that point, there was really this big expectation sort of fueled by
00:10:15.220 --> 00:10:21.460
OpenAI that GPT-5 was going to be the AGI model. And it turns out what was happening internally
00:10:22.080 --> 00:10:27.300
is these scaling laws that were sort of considered, you know, this exponential growth thing that would
00:10:27.620 --> 00:10:33.199
sort of push the power of these models perhaps towards human-like performance. They weren't
00:10:33.220 --> 00:10:38.200
laws at all. And of course they started failing. So the model that they had originally pitched as
00:10:38.320 --> 00:10:42.780
GPT-4 just didn't live up to performance. They started this post-training stuff where they were
00:10:43.200 --> 00:10:48.160
going more into like specialized reasoning models. And what we have now are good models that are good
00:10:48.300 --> 00:10:54.400
for specific tasks, but I don't know what happened, but eventually they had to put the GPT-5 label on
00:10:54.560 --> 00:11:01.959
something. And yeah, let's say it didn't live up to expectations. So I think the cracks are starting
00:11:01.980 --> 00:11:08.620
to show because the underlying expectation always was this will be improving to the point where
00:11:08.960 --> 00:11:14.200
anything's possible and you can't put a price on that. But it turns out that if maybe there's a
00:11:14.320 --> 00:11:19.180
limit on what's possible, yeah, you can put a price on it. And a lot of the valuations are on the
00:11:19.660 --> 00:11:23.780
first part. Yes. And it's always been a bit interesting to me because I come from a scientific
00:11:24.000 --> 00:11:28.459
background and you need to know how to measure stuff, right? And I'm like, what are you trying
00:11:28.480 --> 00:11:34.120
to achieve? Like Gregory's nodding, like, please jump in. I'm on my monologue, so please don't
00:11:34.380 --> 00:11:38.700
interrupt me. You really need to understand what you're actually trying to get these models to do.
00:11:38.920 --> 00:11:46.180
What is AGI? No one knows this. And what's going to be possible with this? And it's more science
00:11:46.420 --> 00:11:51.580
fiction than fact. So this for me has been the big news this year, and I'm feeling slightly smug,
00:11:51.800 --> 00:11:55.180
I'm going to be honest, even though my predictions were off by about a year and a half.
00:11:55.320 --> 00:11:56.720
Yeah, maybe it's not an exponential curve.
00:11:56.920 --> 00:11:59.580
It's a titration S curve with an asymptote.
00:11:59.650 --> 00:11:59.880
We'll see.
00:12:00.020 --> 00:12:01.320
Yeah, sigmoid.
00:12:01.820 --> 00:12:01.940
Yeah.
00:12:02.200 --> 00:12:03.020
Yeah, yeah, yeah.
00:12:03.320 --> 00:12:06.740
I mean, I think we have to sort of separate the technology from the business.
00:12:07.310 --> 00:12:12.780
And the technology, even if it doesn't get any better, even if we stay with what we have today,
00:12:13.260 --> 00:12:17.720
I still think this is like one of the most amazing technologies I've ever seen.
00:12:18.180 --> 00:12:19.200
It's not a god.
00:12:19.520 --> 00:12:20.820
It's not a panacea.
00:12:21.280 --> 00:12:25.200
But it's like a chainsaw that if you know how to use it, it's really effective.
00:12:25.760 --> 00:12:31.980
but in the hands of amateurs, you can really get hurt. And so, yes, it's great to see this sort of
00:12:32.100 --> 00:12:36.260
thing happening and improving, but who knows where it's going to go. And I'm a little skeptical of
00:12:36.260 --> 00:12:40.280
the AGI thing. What I'm a little more worried about is that these companies seem to have no
00:12:40.720 --> 00:12:47.260
possible way of ever making the money that they're promising to their investors. And I do worry a lot
00:12:47.700 --> 00:12:53.580
that we're sort of like a year 2000 situation where, yeah, the technology is fantastic,
00:12:53.660 --> 00:13:03.400
But the businesses are unsustainable. And out of the ashes of what will happen, we will get some amazing technology and even better than we had before. But there are going to be ashes.
00:13:03.520 --> 00:13:09.500
For me, that also makes me worry. And I don't know if anyone reads Ed Zitron here. He's a
00:13:09.750 --> 00:13:15.020
journalist kind of digging into the state of the AI industry. He does get a bit sort of,
00:13:15.340 --> 00:13:19.480
his reputation is a bit of a crank now. So I think he's leaned into that pretty hard,
00:13:20.020 --> 00:13:24.880
but he does take the time to also pull out numbers and point out things that don't make sense.
00:13:25.320 --> 00:13:29.920
And he was one of the first ones to sound the whistle on this circular funding we've been seeing.
00:13:30.160 --> 00:13:40.860
So the worry, of course, is when a lot of this becomes borrowings from banks and then that starts dragging in funding from everyday people.
00:13:41.400 --> 00:13:45.720
And also the effect that this has had on particularly the U.S. economy, like the stock market.
00:13:45.980 --> 00:13:54.380
I think the investment in AI spending now exceeds consumer spending in the U.S., which is a really scary prospect.
00:13:54.860 --> 00:13:55.600
That is crazy.
00:13:55.700 --> 00:14:02.080
Mm-hmm. But yeah, also as Reven said, I love LLMs. They are the most powerful tools we've
00:14:02.090 --> 00:14:06.060
ever had for natural language processing. It's phenomenal the problems we can solve with them
00:14:06.140 --> 00:14:10.100
now. I didn't think this sort of stuff would be possible when I started in data science.
00:14:10.710 --> 00:14:15.560
I still think there's a use case for agents, although I do think they've been a bit overstated,
00:14:16.120 --> 00:14:20.060
especially now that I'm building them. Let's say it's not very fun building
00:14:20.150 -->
00:14:25.099
non-deterministic software. It's quite frustrating, actually. But I hope we're going to see improvements
00:14:25.100 --> 00:14:30.540
in the framework, particularly I've heard good things about Pydantic AI. And yeah, hopefully we
00:14:30.640 --> 00:14:36.080
can control the input outputs and make them a bit more strict. This will fix a lot of the problems.
00:14:36.340 --> 00:14:41.220
One thing I do want to put out in this conversation, I think is worth separating. And Reuven,
00:14:41.220 --> 00:14:45.260
you touched on this some. I want to suggest to you, I'll throw this out to you all and see what
00:14:45.260 --> 00:14:51.599
you think. I think it's very possible that this AI bubble crashes the economy and causes bad things
00:14:51.620 --> 00:14:57.780
economically to happen and a bunch of companies that are like wrappers over open ai api go away
00:14:58.300 --> 00:15:03.800
but i don't think things like the agentic coding tools will vanish they might stop training they
00:15:03.920 --> 00:15:08.880
might slow their advance because that's super expensive but i even as if you said even if we
00:15:09.160 --> 00:15:16.000
just had claude sonnet 4 and the world never got something else it would be so much far farther
00:15:16.150 --> 00:15:20.980
beyond autocomplete and the other stuff that we had before and stack overflow that it's i don't
00:15:20.860 --> 00:15:23.320
think it's going to go. The reason I'm throwing this out there is I was talking to somebody and
00:15:23.340 --> 00:15:26.160
they were like, well, I don't think it's worth learning because I think the bubble is going to
00:15:26.520 --> 00:15:30.020
pop. And so I don't want to learn this agent at coding because it won't be around very long.
00:15:30.340 --> 00:15:34.680
What do you all think? It's here to stay. I think it's just, where's the limit? Where does it stop?
00:15:34.940 --> 00:15:40.340
I think that's the big open question for everybody, right? Like pragmatically, it's a tool. It's useful
00:15:40.410 --> 00:15:44.100
in some scenarios and not in others. And you just have to learn how to use the tool appropriately
00:15:44.300 --> 00:15:47.939
for your use case and to get what you need out of it. And sometimes that's not using it because
00:15:47.960 --> 00:15:51.540
it's just going to take up more time than it will to be productive. But other times it's
00:15:51.960 --> 00:15:56.440
fully juices up your productivity and you can get more done. It's give and take. But I don't think
00:15:56.500 --> 00:16:00.100
it's going to go anywhere because as you said, Michael, there's even academics doing research now.
00:16:00.300 --> 00:16:05.160
There's open weight models as well. There's a lot of different ways to run this, whether you're
00:16:05.520 --> 00:16:11.280
at the scale of the frontier models that are doing these huge trainings or you're doing something
00:16:11.740 --> 00:16:16.140
local and more specialized. So I think the general use of AI isn't going anywhere. I think it's just
00:16:16.080 --> 00:16:21.980
the question of how far can this current trend go and where will it be i want to say stop that's
00:16:22.260 --> 00:16:25.580
because that plays into the whole it's never it's going to completely go away i don't think it ever
00:16:25.680 --> 00:16:29.160
will i think it's just going to be where where are we going to start to potentially bump up against
00:16:29.280 --> 00:16:33.540
limits one thing that i'll say is that many of these systems are almost to me like a dream come
00:16:33.750 --> 00:16:38.640
true now admittedly it's the case that the systems i'm building are maybe only tens of thousands of
00:16:38.700 --> 00:16:44.019
lines or hundreds of thousands of lines but i can remember thinking to myself how cool would it be
00:16:44.080 --> 00:16:46.880
if I had a system that could automatically refactor
00:16:47.440 --> 00:16:51.000
and then add test cases and increase the code coverage
00:16:51.380 --> 00:16:54.140
and make sure all my checkers and linters pass
00:16:54.440 --> 00:16:57.320
and do that automatically and continue the process
00:16:57.650 --> 00:16:58.920
until it achieved its goal.
00:16:59.260 --> 00:17:01.620
And I remember thinking that five to seven years ago,
00:17:01.750 --> 00:17:04.680
I would never realize that goal in my entire lifetime.
00:17:05.189 --> 00:17:07.819
And now when I use like Anthropics models
00:17:08.010 --> 00:17:09.839
through OpenCode or Claude code,
00:17:10.220 --> 00:17:13.120
it's incredible how much you can achieve so quickly,
00:17:13.459 --> 00:17:16.480
even for systems that are of medium to moderate scale.
00:17:17.040 --> 00:17:19.860
So from my vantage point, it is a really exciting tool.
00:17:20.260 --> 00:17:21.420
It's incredibly powerful.
00:17:21.870 --> 00:17:24.339
And what I have found is that the LLMs are much better
00:17:24.640 --> 00:17:26.500
when I teach them how to use tools
00:17:27.040 --> 00:17:28.900
and the tools that it's using
00:17:29.260 --> 00:17:31.220
are actually really quick, fast ones
00:17:31.660 --> 00:17:33.920
that can give rapid feedback to the LLM
00:17:34.240 --> 00:17:35.380
and tell it whether it's moving
00:17:35.450 --> 00:17:36.540
in the right direction or not.
00:17:36.680 --> 00:17:38.900
Yeah, there's an engineering angle to this.
00:17:39.040 --> 00:17:40.180
It's not just Vibe Coding
00:17:40.340 --> 00:17:42.700
if you take the time to learn it.
00:17:42.860 --> 00:17:48.040
There was actually a very interesting study. I don't think the study itself has been released.
00:17:48.060 --> 00:17:53.340
I haven't found it yet, but I saw a talk on it by some guys at Stanford. So they call it the 10K
00:17:53.580 --> 00:17:59.160
developer study. And basically what they were doing was studying real code bases, including,
00:17:59.260 --> 00:18:05.260
I think 80% of them were actually private code bases and seeing the point where the team started
00:18:05.500 -->
00:18:10.699
adopting AI. And so their findings are really interesting and nuanced. And I think they probably
00:18:10.720 --> 00:18:16.080
intuitively align with what a lot of us have experienced with AI. So basically, yes, there
00:18:16.080 --> 00:18:21.720
are productivity boosts, but it produces a lot of code, but the code tends to be worse than the code
00:18:21.720 --> 00:18:27.400
you would write and also introduces more bugs. So when you account for the time that you spend
00:18:27.540 --> 00:18:32.260
refactoring and debugging, you're still more productive. But then it also depends on the
00:18:32.380 --> 00:18:36.500
type of project, as Gregory was saying. So it's better for greenfield projects, it's better for
00:18:36.520 --> 00:18:41.380
smaller code bases. It's better for simpler problems and it's better for more popular languages because
00:18:41.540 --> 00:18:46.200
obviously there's more training data. And so this was actually, I like this study so much. I'll
00:18:46.260 --> 00:18:49.900
actually share it with you, Michael, if you want to put it in the show notes, but it shows that,
00:18:50.000 --> 00:18:53.940
yeah, the picture is not that simple and all this conflicting information and conflicting experiences
00:18:54.180 --> 00:19:00.400
people were having line up completely with this. So again, like I work at an IDE company, it's tools
00:19:00.540 --> 00:19:06.480
for the job. It's not like your IDE will replace you. AI is not going to replace you. It's just
00:19:06.500 --> 00:19:08.200
going to make you maybe more productive sometimes.
00:19:08.700 --> 00:19:08.880
Yeah.
00:19:09.080 --> 00:19:10.400
Wait, IDE, you work for me.
00:19:11.600 --> 00:19:11.700
Right.
00:19:12.500 --> 00:19:13.000
It's not about you.
00:19:13.000 --> 00:19:14.400
But then I work for the IDE.
00:19:18.360 --> 00:19:21.320
This portion of Talk Python To Me is brought to you by Sentry.
00:19:22.640 --> 00:19:23.500
Let me ask you a question.
00:19:24.280 --> 00:19:25.880
What if you could see into the future?
00:19:26.640 --> 00:19:27.900
We're talking about Sentry, of course.
00:19:28.160 --> 00:19:33.460
So that means seeing potential errors, crashes, and bugs before they happen, before you even
00:19:33.780 --> 00:19:34.720
accept them into your code base.
00:19:35.380 --> 00:19:38.500
That's what Sentry's AI Seer Code Review offers.
00:19:39.300 --> 00:19:42.640
You get error prediction based on real production history.
00:19:43.180 --> 00:19:49.620
AI Seer Code Review flags the most impactful errors your PR is likely to introduce before merge
00:19:50.180 --> 00:19:54.980
using your app's error and performance context, not just generic LLM pattern matching.
00:19:55.800 --> 00:20:00.880
Sear will then jump in on new PRs with feedback and warning if it finds any potential issues.
00:20:01.680 --> 00:20:02.460
Here's a real example.
00:20:03.240 --> 00:20:10.560
On a new PR related to a search feature in a web app, we see a comment from seer bicenturybot in the PR.
00:20:11.220 --> 00:20:18.820
And it says, potential bug, the process search results function, can enter an infinite recursion when a search query finds no matches.
00:20:19.620 --> 00:20:23.800
As the recursive call lacks a return statement and a proper termination condition.
00:20:24.360 --> 00:20:32.000
And Seer AI Code Review also provides additional details which you can expand for further information on the issue and suggested fixes.
00:20:32.760 --> 00:20:37.780
And bam, just like that, Seer AI Code Review has stopped a bug in its tracks without any
00:20:37.980 --> 00:20:38.560
devs in the loop.
00:20:38.880 --> 00:20:41.720
A nasty infinite loop bug never made it into production.
00:20:42.440 --> 00:20:43.200
Here's how you set it up.
00:20:43.620 --> 00:20:48.940
You enable the GitHub Sentry integration on your Sentry account, enable Seer AI on your
00:20:49.180 --> 00:20:53.840
Sentry account, and on GitHub, you install the Seer by Sentry app and connect it to your
00:20:53.960 --> 00:20:55.600
repositories that you want it to validate.
00:20:55.940 --> 00:20:58.800
So jump over to Sentry and set up Code Review for yourself.
00:20:59.320 --> 00:21:03.420
Just visit talkpython.fm/seer-code-review.
00:21:03.650 --> 00:21:05.340
The link is in your podcast player show notes
00:21:05.580 --> 00:21:06.460
and on the episode page.
00:21:06.980 --> 00:21:09.260
Thank you to Sentry for supporting Talk Python and me.
00:21:10.140 --> 00:21:12.000
I mean, the other thing is a lot of people
00:21:12.440 --> 00:21:15.060
and a lot of the sort of when people talk about AI
00:21:15.400 --> 00:21:17.280
and LLMs and so forth in context of coding,
00:21:17.800 --> 00:21:20.500
it's the LLM writing code for us.
00:21:20.920 --> 00:21:23.440
And maybe because I'm not doing a lot of serious coding,
00:21:23.600 --> 00:21:25.060
it's more instruction and so forth.
00:21:25.320 --> 00:21:29.280
I use it as like a sparring or brainstorming partner
00:21:29.620 --> 00:21:36.800
So it does, you know, checking of my newsletters for language and for tech edits and just sort of exploring ideas.
00:21:37.300 --> 00:21:43.320
And for that, maybe it's because I do everything in the last minute and I don't have other people around or I'm lazy or cheap and don't want to pay them.
00:21:43.800 --> 00:21:46.620
But definitely the quality of my work has improved dramatically.
00:21:46.860 --> 00:21:50.680
The quality of my understanding has improved, even if it never wrote a line of code for me.
00:21:50.920 --> 00:21:53.860
Just getting that feedback on a regular automatic basis is really helpful.
00:21:54.120 --> 00:21:55.140
Yeah, I totally agree with you.
00:21:55.360 --> 00:22:05.520
All right. We don't want to spend too much time on this topic, even though I believe Jody has put her finger on what might be the biggest tidal wave of 2025.
00:22:06.040 --> 00:22:08.560
But still, a quick parting thoughts. Anyone else?
00:22:08.900 -->
00:22:11.220
I'm glad I'll never have the right bash from scratch ever again.
00:22:12.960 --> 00:22:13.640
Tell me about it.
00:22:13.740 --> 00:22:14.040
Yeah.
00:22:15.800 --> 00:22:21.580
I'll just say from anecdotally, the thing that I love about it is when I need to do something
00:22:22.020 --> 00:22:28.220
and I need to go through docs, online docs for whatever it is, you know, it might be
00:22:28.380 --> 00:22:31.900
GitLab or some library that I want to use or something like that.
00:22:32.150 --> 00:22:33.800
I never even search for the docs.
00:22:33.810 --> 00:22:35.680
I never even try to read the docs anymore.
00:22:35.810 --> 00:22:40.940
I just say, hey, you know, whatever model I need to set up this website.
00:22:41.300 --> 00:22:43.720
And I just, just tell me what to do or just do it.
00:22:43.960 --> 00:22:47.460
And it's an immense time saver and productivity.
00:22:47.680 --> 00:22:52.120
And then it gets me bootstrapped to the place where now I can start to be creative.
00:22:52.350 --> 00:22:57.800
I don't have to worry about just like digging through pages and pages and pages of docs to
00:22:57.950 --> 00:23:00.000
figure out one little setting here or there.
00:23:00.320 --> 00:23:02.040
That's an amazing time saver.
00:23:02.120 --> 00:23:03.220
Yeah, that's a really good point.
00:23:03.500 --> 00:23:07.160
Another thing that I have noticed, there might be many things for which I had a really good
00:23:07.390 --> 00:23:10.420
mental model, but my brain can only store so much information.
00:23:10.960 --> 00:23:16.000
So for example, I know lots about the abstract syntax tree for Python, but I forget that
00:23:16.150 --> 00:23:16.280
sometimes.
00:23:16.870 --> 00:23:21.280
And so it's really nice for me to be able to bring that back into my mind quickly with
00:23:21.280 --> 00:23:21.960
an LLM.
00:23:22.250 --> 00:23:26.940
And if it's generating code for me that's doing a type of AST parsing, I can tell whether
00:23:27.080 --> 00:23:30.380
that's good code or not because I can refresh that mental model.
00:23:30.740 --> 00:23:34.960
So in those situations, it's not only the docs, but it's something that I used to know
00:23:35.070 --> 00:23:37.600
really well that I have forgotten some of.
00:23:37.940 --> 00:23:44.180
And the LLM often is very powerful when it comes to refreshing my memory and helping me to get started and move more quickly.
00:23:44.320 --> 00:23:48.680
All right. Out of time, I think. Let's move on to Brett. What do you got, Brett?
00:23:48.920 --> 00:23:56.020
Well, I actually originally said we should talk about AI, but Jody had a way better pitch for it than I did because my internal pitch was a little bit AI.
00:23:56.360 --> 00:24:01.720
Do I actually have to write a paragraph explaining why? Then Jody actually did write the paragraph. So she did a much better job than I did.
00:24:01.940 --> 00:24:06.320
So the other topic I had was using tools to run your Python code.
00:24:06.680 --> 00:24:09.060
And what I mean by that is traditionally, if you think about it,
00:24:10.120 --> 00:24:11.760
you install the Python interpreter, right?
00:24:13.480 --> 00:24:15.840
Hopefully you create a virtual environment, install your dependencies,
00:24:16.140 --> 00:24:18.940
and you call the Python interpreter in your virtual environment to run your code.
00:24:19.560 --> 00:24:21.020
Those are all the steps you went through to run stuff.
00:24:21.360 --> 00:24:25.460
But now we've got tools that will compress all that into a run command,
00:24:25.920 --> 00:24:26.640
just do it all for you.
00:24:26.960 --> 00:24:31.520
And it seems like the community has shown a level of comfort with that,
00:24:31.940 --> 00:24:34.380
that I'd say snuck up on me a little bit,
00:24:34.680 --> 00:24:37.620
but I would say that I think it's a good thing, right?
00:24:37.780 --> 00:24:40.200
It's showing us, I'm going to say us,
00:24:40.370 --> 00:24:43.060
as the junior core developer here on this call,
00:24:43.440 --> 00:24:45.700
as to, sorry to make you too feel old,
00:24:45.840 --> 00:24:48.680
but admittedly, Barry did write my letter of recommendation
00:24:48.910 --> 00:24:49.940
to my master's program.
00:24:51.260 --> 00:24:52.800
So what happened was like, yeah,
00:24:53.100 --> 00:24:54.740
we had Hatch and PDM,
00:24:55.290 --> 00:24:58.180
poetry before that, and uv as of last year,
00:24:58.680 --> 00:24:59.360
all kind of come through
00:24:59.720 --> 00:25:00.880
and all kind of build on each other
00:25:01.200 --> 00:25:02.100
and take ideas from each other
00:25:02.240 --> 00:25:03.680
and kind of just slowly build up
00:25:03.680 --> 00:25:05.940
this kind of repertoire of tool approaches
00:25:06.260 --> 00:25:08.580
that they all kind of have a baseline kind of,
00:25:08.900 --> 00:25:09.740
not synergy is the right word,
00:25:09.960 --> 00:25:12.720
but share just kind of approach to certain things
00:25:13.220 --> 00:25:15.640
with their own twists and added takes on things.
00:25:15.790 --> 00:25:17.300
But in general, this whole like,
00:25:17.580 --> 00:25:19.300
you know what, you can just tell us to run this code
00:25:19.400 --> 00:25:20.440
and we will just run it, right?
00:25:20.540 --> 00:25:22.160
Like inline script metadata coming in
00:25:22.260 --> 00:25:23.980
and help making that more of a thing.
00:25:24.380 --> 00:25:26.860
Disclaimer, I was the PEP delegate for getting that in.
00:25:27.140 --> 00:25:29.400
But I just think that's been a really awesome trend
00:25:29.420 --> 00:25:33.900
And I'm hoping we can kind of leverage that a bit.
00:25:34.190 --> 00:25:36.660
Like I have personal plans that we don't need to go into here,
00:25:36.820 --> 00:25:38.820
but like I'm hoping as a Python core team,
00:25:38.870 --> 00:25:41.020
we can kind of like help boost this stuff up a bit
00:25:41.020 --> 00:25:43.300
and kind of help keep a good baseline for this for everyone.
00:25:43.350 --> 00:25:45.700
Because
I think it's shown that Python is still really good for beginners.
00:25:45.970 --> 00:25:48.900
You just have to give them the tools to kind of hide some of the details
00:25:49.280 --> 00:25:51.800
to not shoot yourself in the foot and still leads to a great outcome.
00:25:52.000 --> 00:25:55.640
Yeah, 2025 might be the year that the Python tools stepped outside of Python.
00:25:56.180 --> 00:25:59.740
Instead of being, you install Python and then use the tools.
00:26:00.190 --> 00:26:01.920
You do the tool to get Python, right?
00:26:02.140 --> 00:26:03.660
Like uv and PDM and others.
00:26:03.670 --> 00:26:07.880
Yeah, and inverted the dependency graph in terms of just how you put yourself in, right?
00:26:08.040 --> 00:26:13.260
I think the interesting thing is these tools treat Python as an implementation detail almost, right?
00:26:13.520 --> 00:26:17.480
Like when you just say uv or hatch run or PDM run thing,
00:26:17.920 --> 00:26:19.520
these tools don't make you have to think about the interpreter.
00:26:19.740 --> 00:26:22.480
It's just a thing that they pull in to make your code run, right?
00:26:22.600 --> 00:26:25.940
It's not even necessarily something you have to care about if you choose not to.
00:26:26.100 --> 00:26:29.940
And it's an interesting shift in that perspective, at least for me.
00:26:30.050 --> 00:26:31.360
But I've also been doing this for a long time.
00:26:31.360 --> 00:26:33.080
I think you're really onto something.
00:26:33.460 --> 00:26:39.480
And what I love at sort of a high level is this, I think there's a renewed focus on the user experience.
00:26:40.320 --> 00:26:49.080
And like uv plus the PEP 723, the inline metadata, you know, you can put uv in the shebang line of your script.
00:26:49.560 --> 00:26:51.760
And now you don't have to think about anything.
00:26:52.120 --> 00:26:56.340
You get uv from somewhere, and then it takes care of everything.
00:26:57.790 --> 00:27:00.800
And Hatch can work the same way, I think, for developers.
00:27:01.100 --> 00:27:08.140
But this renewed focus on installing your Python executable,
00:27:08.340 --> 00:27:11.880
you don't really have to think about, because those things are very complicated,
00:27:12.030 --> 00:27:14.420
and people just want to hit the ground running.
00:27:14.950 --> 00:27:17.600
And so if you think about the previous discussion about AI,
00:27:18.260 --> 00:27:19.880
I just want things to work.
00:27:20.140 --> 00:27:21.760
I know what I want to do.
00:27:22.220 --> 00:27:23.140
I can see it.
00:27:23.140 --> 00:27:24.300
I can see the vision of it.
00:27:24.300 --> 00:27:25.640
And I just don't want to.
00:27:25.960 --> 00:27:28.740
An analogy is like when I first learned Python
00:27:28.960 --> 00:27:31.580
and I came from C++ and all those languages.
00:27:32.280 --> 00:27:35.140
And I thought, oh my gosh, just to get like, hello world,
00:27:35.280 --> 00:27:39.420
I have to do a million little things that I shouldn't have to do.
00:27:39.680 --> 00:27:43.120
Like create a main and get my braces right
00:27:43.140 --> 00:27:46.760
and get all my variables right and get my pound includes correct.
00:27:46.920 --> 00:27:48.980
And now I don't have to think about any of that stuff.
00:27:49.260 --> 00:27:59.980
And the thing that was eye-opening for me with Python was the distance between vision of what I wanted and working code just really narrowed.
00:28:00.100 --> 00:28:09.320
And I think that as we are starting to think about tools and environments and how to bootstrap all this stuff, we're also now taking all that stuff away.
00:28:09.840 --> 00:28:11.580
Because people honestly don't care.
00:28:11.650 --> 00:28:13.320
I don't care about any of that stuff.
00:28:13.330 --> 00:28:18.020
I just want to go from like, I woke up this morning and had a cool idea and I just wanted to get at work.
00:28:18.860 --> 00:28:22.080
Or you wanted to share it so you could just share the script and you don't have to say,
00:28:22.460 --> 00:28:24.300
here's your steps that you get started with.
00:28:24.980 --> 00:28:25.300
Exactly.
00:28:25.800 --> 00:28:26.020
Exactly.
00:28:26.320 --> 00:28:28.660
I want to thank the two of you for, oh, sorry, sorry, go ahead.
00:28:28.960 --> 00:28:34.220
I'm just going to say, like, for years teaching Python that how do we get it installed?
00:28:34.720 --> 00:28:37.580
At first, it surprised me how difficult it was for people.
00:28:37.700 --> 00:28:39.260
Because like, oh, come on, we just got Python.
00:28:39.460 --> 00:28:40.380
Like, what's so hard about this?
00:28:40.620 --> 00:28:44.740
But it turns out it's a really big barrier to entry for newcomers.
00:28:45.200 --> 00:28:48.920
And I'm very happy that JupyterLite now has solved its problems with input.
00:28:49.320 --> 00:28:50.520
And it's like huge.
00:28:50.960 --> 00:28:56.820
But until now, I hadn't really thought about starting with uv because it's cross-platform.
00:28:57.180 --> 00:29:02.100
And if I say to people in the first 10 minutes of class, install uv for your platform and
00:29:02.100 --> 00:29:05.100
then say uv in it, your project, bam, you're done.
00:29:05.320 --> 00:29:06.060
It just works.
00:29:06.440 --> 00:29:07.200
And then it works cross-platform.
00:29:07.580 --> 00:29:08.520
This is mind-blowing.
00:29:08.700 --> 00:29:10.140
And I'm going to try this at some point.
00:29:10.220 --> 00:29:10.540
Thank you.
00:29:10.640 --> 00:29:14.860
I can comment on the mind-blowing part because now when I teach undergraduate students, we
00:29:14.880 --> 00:29:20.360
start with uv in the very first class. And it is awesome. There were things that would take students,
00:29:20.820 --> 00:29:25.920
even very strong students who've had lots of experience, it would still take them a week to
00:29:26.160 --> 00:29:30.160
set everything up on their new laptop and get everything ready and to understand all the key
00:29:30.580 --> 00:29:36.640
concepts and know where
something is in their path. And now we just say, install uv for your
00:29:36.900 --> 00:29:42.820
operating system and get running on your computer. And then, hey, you're ready to go. And I don't have
00:29:42.780 --> 00:29:47.040
teach them about docker containers and i don't have to tell them how to install python with some
00:29:47.160 --> 00:29:52.900
package manager all of those things just work and i think from a learning perspective whether you're
00:29:52.960 --> 00:29:58.540
in a class or whether you're in a company or whether you're teaching yourself uv is absolutely
00:29:58.900 --> 00:30:04.960
awesome i'm actually wondering whether i am the one who is newest to python here i taught myself
00:30:05.280 --> 00:30:12.740
python in 2011 so i was like python 2.7 stage but it was my first programming language i was just
00:30:12.760 --> 00:30:17.920
procrastinating during my PhD. And I was like, I should learn to program. So I just taught myself
00:30:18.140 --> 00:30:23.360
Python. And I can tell you, you do not come from an engineering background. And you're like,
00:30:23.430 --> 00:30:29.240
what is Python? What is Python doing? Why am I typing Python to execute this hello world? And
00:30:29.300 --> 00:30:33.840
if you're kind of curious, you get down a rabbit hole before you even get to the point where you're
00:30:33.960 --> 00:30:39.220
just focusing on learning the basics. And so it's exactly, I was going to say with Reuven,
00:30:39.240 --> 00:30:43.160
And like whether you thought about it for teaching, because we're now debating for Humble Data,
00:30:43.260 --> 00:30:48.040
which is a beginner's data science community that I'm part of, whether we switch to uv.
00:30:48.320 --> 00:30:52.180
This was Chuck's idea because it does abstract away all these details.
00:30:52.720 --> 00:30:55.320
The debate I have is, is it too magic?
00:30:55.820 --> 00:30:59.200
This is kind of the problem because I also remember learning about things like virtual
00:30:59.500 --> 00:31:02.540
environments, because again, this was my first programming language and being like,
00:31:02.780 --> 00:31:03.740
oh, it's a very good idea.
00:31:04.060 --> 00:31:04.980
This is best practices.
00:31:05.280 --> 00:31:07.920
And it's also a debate we have in PyCharm, right?
00:31:08.220 --> 00:31:14.780
Like how much do you magic away the fundamentals versus making people think a little bit, but
00:31:15.140 --> 00:31:15.560
I'm not sure.
00:31:15.660 --> 00:31:15.840
All right.
00:31:15.920 --> 00:31:18.720
Like, would you even let somebody run without a virtual environment?
00:31:19.000 --> 00:31:21.180
That's like, that's a take you, that's a stance you could take.
00:31:21.300 --> 00:31:27.620
I used to when I first learned Python, because it was too complicated, but then I learned
00:31:27.840 --> 00:31:28.140
better.
00:31:28.640 --> 00:31:29.160
But yes.
00:31:29.320 --> 00:31:34.380
The consideration here is like hiding the magic isn't like hiding the
details and having
00:31:34.430 --> 00:31:37.580
all this magic just work is great as long as it works.
00:31:38.020 --> 00:31:43.620
And the question is, how is it going to break down and how are people going to know how to deal
00:31:43.830 --> 00:31:49.540
when it breaks down, if you hide all the magic? And I think virtual envs were, or let's say before
00:31:49.650 --> 00:31:54.740
we had virtual envs, installing packages was very much in the, you had to know all the details
00:31:54.910 --> 00:32:00.300
because it was very likely going to break down in some way right before we had virtual envs,
00:32:00.560 --> 00:32:04.880
because you would end up with weird conflicts or multiple copies of a package installed in
00:32:04.840 --> 00:32:09.820
different parts of the system. When we got virtual ends, we sort of didn't have to worry about that
00:32:10.020 --> 00:32:14.380
anymore because we were trained in that you can just blow away the virtual one and it just works.
00:32:14.920 --> 00:32:19.900
And with uv, we're back into, this looks like a single installation. We don't know what's going
00:32:19.900 --> 00:32:25.360
to go on, but we've learned, we as a community and also the people working on uv, we have learned
00:32:25.700 --> 00:32:31.640
from those earlier mistakes or not, maybe not mistakes, but consequences of the design.
00:32:32.180 --> 00:32:38.160
And they have created something that is, that appears to be very stable where it's unlikely
00:32:38.280 --> 00:32:39.220
the magic will break.
00:32:39.460 --> 00:32:43.940
And when the magic does break, it's obvious what the problem is or, or it automatically
00:32:44.100 --> 00:32:44.620
fixes itself.
00:32:45.100 --> 00:32:50.480
So like it's not reusing, broken, installations and that kind of thing.
00:32:50.720 --> 00:32:57.220
So the risk now, as it turns out, I think as is proven by the community adopting uv so
00:32:57.860 --> 00:33:00.160
fast and so willingly, I think it's acceptable.
00:33:00.180 --> 00:33:01.860
Well, I think it's, yeah, I think it's proven itself.
00:33:02.140 --> 00:33:09.380
It's clear that this is, it's worth the potential of discovering weird edge cases later, both
00:33:09.520 --> 00:33:16.060
because it's probably low likelihood, but also the people behind uv Astral have proven that
00:33:16.200 --> 00:33:18.600
they would jump in and fix those issues, right?
00:33:18.600 --> 00:33:22.840
They would do anything they need to keep uv workable the same way.
00:33:23.190 --> 00:33:28.920
And they have a focus that Python as a whole cannot have because they cater to fewer use
00:33:28.940 --> 00:33:31.200
cases than Python as a whole needs to.
00:33:31.320 --> 00:33:36.160
On the audience, Galano says, as an enterprise tech director in Python coder, I believe we
00:33:36.300 --> 00:33:40.500
should hide the magic which empowers the regular employee to do simple things that make their
00:33:40.620 --> 00:33:40.960
job easier.
00:33:41.420 --> 00:33:41.500
Yeah.
00:33:41.650 --> 00:33:46.800
This notion of abstractions, right, has always been there in computer science.
00:33:47.980 --> 00:33:53.560
And, you know, we've used tools or languages or systems where we've tried to bring that
00:33:53.580 --> 00:33:58.600
abstraction layer up so that we don't have to think about all these details, as I mentioned
00:33:58.790 --> 00:34:04.080
before. The question is, that's always the happy path. And when I'm trying to teach somebody
00:34:04.300 --> 00:34:08.960
something like, here's how to use this library or here's how to use this tool, I try to be very
00:34:09.220 --> 00:34:14.659
opinionated to keep people on that happy path. Like, assume everything's going to work just right.
00:34:15.020 --> 00:34:19.880
Here's how you just make you go down that path to get the thing done that you want. The question
00:34:19.899 --> 00:34:27.200
really is when things go wrong, how narrow is that abstraction? And are you able, and even when
00:34:27.200 --> 00:34:31.960
you're just curious, like what's really going on underneath the hood? Of course, that's not a really
00:34:31.960 --> 00:34:36.540
good analogy today because cars are basically computers on wheels that you can't really
00:34:36.760 --> 00:34:42.540
understand how they work. But back in your day. But back in my day, we were changing spark plugs,
00:34:42.620 --> 00:34:49.860
you know, but and crank that window down. Exactly. So I think we always have to leave that
00:34:50.280 --> 00:34:56.620
room for the curious and the bad path where when things go wrong or when you're just like,
00:34:56.810 --> 00:35:01.540
you know what, I understand how this works, but I'm kind of curious about what's really going on.
00:35:01.780 --> 00:35:06.920
How easy is it for me to dive in and get a little bit more of that background, you know,
00:35:07.140 --> 00:35:11.620
a little bit more of that understanding of what's going on. I want the magic to decompose,
00:35:12.020 --> 00:35:16.880
right like you should be able to explain the magic path via a more decomposed steps using the tool
00:35:17.080 --> 00:35:21.760
all the way down to what the tools like to do behind the scenes just just to admit i the reason
00:35:21.850 --> 00:35:25.900
i brought this up and i've been thinking about this a lot is i'm thinking of trying to get the python
00:35:26.140 --> 00:35:31.380
launcher to do a bit more because one interesting thing we haven't really brought up here is we're
00:35:31.390 --> 00:35:36.960
all seeing uv uv uv uv is a company there's always there's they might disappear and we haven't
00:35:36.980 --> 00:35:41.640
de-risked ourselves from that. Now we do have Hatch, we do have PDM, but as I said, there's kind
00:35:41.640 --> 00:35:45.440
of a baseline I think they all share that I think they would be probably okay if the Python launcher
00:35:45.580 --> 00:35:48.800
just did because that's based on standards, right? Because that's the other thing that there's been
00:35:48.800 --> 00:35:52.600
a lot of work that has led to this step, right? Like we've gotten way more packaging standards,
00:35:52.900 --> 00:35:58.220
we've got PEP 723, like we mentioned. There's a lot of work that's come up to lead to this point
00:35:58.460 --> 00:36:03.920
that all these tools can lean on to have them all have an equivalent outcome because it's expected
00:36:03.940 --> 00:36:04.700
is how they should be.
00:36:05.199 --> 00:36:07.740
And so I think it's something we need to consider
00:36:08.210 --> 00:36:09.360
of how do we make sure,
00:36:09.820 --> 00:36:10.740
like, by the way, uv,
00:36:11.230 --> 00:36:12.140
I know the people, they're great.
00:36:12.280 --> 00:36:13.640
I'm not trying to spares them
00:36:13.760 --> 00:36:14.620
or think they're going to go away,
00:36:14.780 --> 00:36:16.320
but it is something we have to consider.
00:36:16.800 --> 00:36:18.100
And I will also say, Jody,
00:36:18.480 --> 00:36:19.780
I do think about this for teaching
00:36:20.080 --> 00:36:21.480
because I'm a dad now
00:36:21.880 --> 00:36:23.460
and I don't want my kid coming home
00:36:24.180 --> 00:36:25.820
when they get old enough to learn Python
00:36:25.910 --> 00:36:26.720
and go, hey, dad,
00:36:26.880 --> 00:36:28.980
why is getting Python code running so hard?
00:36:29.240 --> 00:36:31.420
So I want to make sure that that never happens.
00:36:32.160 --> 00:36:34.680
But they fall in love with it from the start.
00:36:34.940 --> 00:36:37.600
I realized something for the 2026 year interview.
00:36:37.680 --> 00:36:42.960
I have to bring a sign that says time for next topic because we got a bunch of topics and we're running low on time.
00:36:43.640 --> 00:36:45.660
So, Thomas, let's jump over to yours.
00:36:46.120 --> 00:36:48.240
Oh, and I had two topics as well.
00:36:48.400 --> 00:36:52.340
So I'm only going to have to pick my favorite child, right?
00:36:52.540 --> 00:36:52.920
That's terrible.
00:36:53.900 --> 00:36:57.960
My second favorite child is Lazy Imports, which is a relatively new development.
00:36:58.280 --> 00:37:00.000
So we'll probably not get to that.
00:37:00.020 --> 00:37:00.120
And just accepted.
00:37:00.300 --> 00:37:02.360
Yes, it's been accepted and it's going to be awesome.
00:37:03.140 --> 00:37:06.620
So I'll just give that a shout out and then move to my favorite child, which is free threaded
00:37:06.850 --> 00:37:06.980
Python.
00:37:07.580 --> 00:37:11.640
For those who were not aware, the global interpreter lock is going away.
00:37:12.160 --> 00:37:13.700
I am stating it as a fact.
00:37:13.840 --> 00:37:17.780
It's not actually a fact yet, but it, you know, that's because the steering council hasn't
00:37:18.240 --> 00:37:19.460
realized the fact yet.
00:37:20.860 --> 00:37:21.940
It is trending towards.
00:37:22.320 --> 00:37:28.820
Well, I was originally on the steering council that accepted the proposal to add free threading
00:37:28.840 --> 00:37:34.500
as a, as an
experimental feature, we had this idea of adding it as experimental and then making it
00:37:34.600 --> 00:37:39.300
supported, but not the default and then making it the default. And it was all a little vague and,
00:37:39.500 --> 00:37:43.800
and up in the air. And then I didn't get reelected for the steering council last year,
00:37:44.060 --> 00:37:49.420
which I was not sad about at all. I sort of ran on a, well, if there's nobody better, I'll do it,
00:37:49.520 --> 00:37:54.820
but otherwise I have other things to do. And it turns out those other things were making sure that
00:37:54.840 --> 00:37:59.700
prefer that Python landed in a supported state. So I lobbied the steering council quite hard,
00:38:00.040 --> 00:38:04.740
as Barry might remember at the start of the year, to get some movement on this, like get some
00:38:04.900 --> 00:38:10.500
decision going. So for Python 3.14, it is officially supported. The performance is great. It's like
00:38:10.820 --> 00:38:16.360
between a couple of percent slower and 10% slower, depending on the hardware and the compiler that
00:38:16.360 --> 00:38:23.240
you use. It's basically the same speed on macOS, which is really like it's, that's a combination of
00:38:23.380 --> 00:38:29.240
the ARM hardware and Clang specializing things, but it's basically the same speed, which, wow.
00:38:29.580 --> 00:38:35.040
And then on recent GCCs on Linux, it's like a couple of percent slower. The main problem is
00:38:35.080 --> 00:38:40.380
really community adoption, getting third-party packages to update their extension modules for
00:38:40.780 --> 00:38:48.560
the new APIs and the things that by necessity sort of broke, and also supporting free threading in a
00:38:48.420 --> 00:38:54.060
in a good way and in packages for Python code, it turns out there's very few changes that
00:38:54.070 --> 00:38:56.840
need to be made for things to work well under free threading.
00:38:57.400 --> 00:39:02.680
They might not be entirely thread safe, but usually like almost always in cases where it
00:39:02.920 --> 00:39:07.080
wasn't thread saved before either, because the guild doesn't actually affect thread safety.
00:39:07.820 --> 00:39:09.300
Just the likelihood of things breaking.
00:39:09.600 --> 00:39:14.920
I do think there's been a bit of a, the mindset of the Python community hasn't really been
00:39:14.940 --> 00:39:18.380
focused on creating thread safe code because the GIL is supposed to protect us.
00:39:18.460 --> 00:39:22.160
But soon as it takes multiple steps, then all of a sudden it's just less likely.
00:39:22.340 --> 00:39:23.540
It's not that it couldn't happen.
00:39:23.780 --> 00:39:24.760
Yeah, that's my point, right?
00:39:24.860 --> 00:39:27.160
It's not the GIL never gave you threat safety.
00:39:27.460 --> 00:39:30.720
The GIL gave cpythons internals threat safety.
00:39:31.120 --> 00:39:35.980
It never really affected Python code and it very rarely affected thread safety in
00:39:36.260 -->
00:39:37.480
extension modules as well.
00:39:37.620 --> 00:39:41.780
So they already had to take care of, of making sure that the global interpreter
00:39:41.800 --> 00:39:46.560
couldn't be released by something that they ended up calling indirectly so it's actually not that
00:39:46.760 --> 00:39:52.980
hard to port most things to support free threading and the benefits we've seen some experimental
00:39:53.030 --> 00:39:56.320
work because you know it's still it's still new there's still a lot of things that don't
00:39:56.720 --> 00:40:02.060
quite support it there's still places where thread contention slows things things down a lot but
00:40:02.180 --> 00:40:10.020
we've seen a lot of examples of really promising very parallel problems that now speed up by 10x or
00:40:09.980 --> 00:40:15.560
more. And it's going to be really excited in the future. And it's in 2025 that this all started.
00:40:15.820 --> 00:40:21.620
I mean, Sam started it earlier, but he's been working on this for years, but it landed in 2025.
00:40:21.880 --> 00:40:26.760
It dropped its experimental stage in 314, basically. Yeah. I was going to say, were we all,
00:40:27.040 --> 00:40:30.760
the three of us on the steering council at the same time when we decided to start the experiment
00:40:30.940 --> 00:40:36.320
for free threading? I think Barry wasn't on it. Yeah, I missed a couple of years there, but I'm
00:40:36.340 --> 00:40:36.720
Not sure.
00:40:36.920 --> 00:40:37.640
No, I totally agree.
00:40:37.640 --> 00:40:48.600
I think free threading is one of the most transformative developments for Python, certainly since Python 3, but even maybe more impactful because of the size of the community today.
00:40:49.300 --> 00:40:56.420
Personally, you know, not necessarily speaking as a current or potentially former steering council member.
00:40:56.680 --> 00:40:58.480
We'll see how that shakes out.
00:40:58.720 --> 00:41:00.080
But I think it's inevitable.
00:41:00.460 --> 00:41:08.360
I think free threading is absolutely the future of Python, and I think it's going to unlock incredible potential and performance.
00:41:08.930 --> 00:41:10.460
I think we just have to do it right.
00:41:10.630 --> 00:41:15.800
And so I talked to lots of teams who are building various software all over the community.
00:41:16.640 --> 00:41:23.580
And I actually think it's more of an educational and maybe an outreach problem than it is a technological problem.
00:41:23.720 --> 00:41:28.800
I mean, yes, there are probably APIs that are missing that will make people's lives easier.
00:41:30.300 --> 00:41:36.040
There's probably some libraries that will make other code a little easier to write or whatever or to understand.
00:41:36.900 --> 00:41:37.980
But like all that's solvable.
00:41:38.100 --> 00:41:42.620
And I think really reaching out to the teams that are, you know, like Thomas said,
00:41:42.640 --> 00:41:46.860
that are building the ecosystem, that are moving the ecosystem to a free threading world.
00:41:47.380 --> 00:41:50.040
That's where we really need to spend our effort on.
00:41:50.160 --> 00:41:51.100
And we'll get there.
00:41:51.360 --> 00:41:52.140
It won't be that long.
00:41:52.400 --> 00:41:55.480
It certainly won't be as long as it took us to get to Python 3.
00:41:56.920 --> 00:42:04.000
I'm sort of curious as someone who's not super experienced with threading or, you know, basic concurrency.
00:42:04.560 --> 00:42:13.820
I mean, I've used it, but I feel like now we have threads, especially with free threading and sub interpreters and multiprocessing and asyncio.
00:42:14.360 --> 00:42:19.560
And I feel like for many people now it's like, oh, my God, which one am I supposed to use?
00:42:19.920 --> 00:42:24.020
And for someone who's experienced, you can sort of say, well, this seems like a better choice.
00:42:24.140 --> 00:42:31.060
But are there any plans to sort of try to have a taxonomy of what problems are solved by which of these?
00:42:31.380 --> 00:42:37.360
The premise here is that everyone would be using one or more of these low-level techniques that you mentioned.
00:42:37.700 --> 00:42:40.620
And I think that's not a good way of looking at it.
00:42:40.860 --> 00:42:46.340
Like AsyncIO is a library that you want to use for the things that AsyncIO is good at.
00:42:46.580 --> 00:42:52.220
And you can actually very nicely combine it with multiprocessing, with subprocesses, with
00:42:52.510 --> 00:42:57.700
so that subprocesses and subinterpreters, just to make it clear that those are two very separate
00:42:57.860 --> 00:43:00.800
things and multithreading, both with and without free threading.
00:43:01.100 --> 00:43:06.560
And it solves different problems or it gives you different abilities within the AsyncIO
00:43:06.760 --> 00:43:06.900
framework.
00:43:06.950 --> 00:43:09.100
And the same is true for like GUI frameworks.
00:43:09.730 --> 00:43:14.360
I mean, GUI frameworks usually want threads for multiple reasons, but you can use these
00:43:14.490 --> 00:43:15.300
other things as well.
00:43:15.720 --> 00:43:21.580
I don't think it's down to teaching end users when to use or avoid all these different things.
00:43:22.070 --> 00:43:26.960
I think we need higher level abstractions for tasks that people want to solve.
00:43:27.270 --> 00:43:32.980
And then those can decide on what for their particular use case is a better approach.
00:43:33.170 --> 00:43:35.840
For instance, PyTorch has multiple.
00:43:36.090 --> 00:43:42.180
So it's used for people who don't know to train, not just train, but it's used in AI
00:43:42.200 --> 00:43:45.720
for generating large matrices and LLMs and what have you.
00:43:46.100 --> 00:43:49.160
Part of it is loading data and processing.
00:43:49.680 --> 00:43:52.640
And the basic ideas of AsyncIO are,
00:43:52.860 --> 00:43:54.720
oh, you can do all these things in parallel
00:43:55.040 --> 00:43:56.500
because you're not waiting on the CPU,
00:43:56.720 -->
00:43:57.680
you're just waiting on IO.
00:43:58.200 --> 00:44:00.720
Turns out it is still a good idea to use threads
00:44:00.960 --> 00:44:02.400
for massively parallel IO
00:44:02.600 --> 00:44:05.740
because otherwise you end up waiting longer than you need to.
00:44:05.940 --> 00:44:10.120
So a problem where we thought AsyncIO would be the solution
00:44:10.140 --> 00:44:15.520
and we never needed threads is actually much improved if we tie in threads as well.
00:44:15.730 --> 00:44:19.200
And we've seen massive, massive improvements in data loader.
00:44:19.320 --> 00:44:23.660
There's even an article, a published article from some people at Meta
00:44:24.200 --> 00:44:28.840
showing how much they improve the PyTorch data loader by using multiple threads.
00:44:29.210 --> 00:44:33.240
But at a very low level, we don't want end users to need to make that choice, right?
00:44:33.300 --> 00:44:35.380
I concur to futures is a good point, right?
00:44:35.500 --> 00:44:39.300
Like all of these approaches are all supported there and it's a unified one.
00:44:39.880 --> 00:44:42.320
So if you were to teach this, for instance, you could say use concurrent.futures.
00:44:42.880 --> 00:44:43.500
These are all there.
00:44:44.130 --> 00:44:45.460
This is the potential tradeoff.
00:44:45.550 --> 00:44:46.800
Like basically use threads.
00:44:46.940 --> 00:44:51.280
It's going to be the fastest unless there's like some module you have that's not that's
00:44:51.400 --> 00:44:53.040
screwing up because of threads, then use sub interpreters.
00:44:53.540 --> 00:44:57.540
And if for some reason sub interpreters don't work, you should move to the processing pool,
00:44:57.710 --> 00:44:58.340
the process pool.
00:44:58.680 --> 00:45:02.140
But I mean, basically, you just kind of just like, it's not go sort the fast stuff.
00:45:02.190 --> 00:45:03.300
And for some reason, it doesn't work.
00:45:03.390 --> 00:45:05.420
Use the next fastest and just kind of do it that way.
00:45:05.860 --> 00:45:07.820
After that, then you start to the lower level.
00:45:08.240 --> 00:45:11.240
Like, okay, why do I want to use subinterpreters instead of threads?
00:45:11.730 --> 00:45:12.260
Those kinds of threads.
00:45:12.390 --> 00:45:15.600
But I think that's a different, as I think we're all searching,
00:45:15.830 --> 00:45:18.840
a different level of abstraction, which is a term we keep bringing up today.
00:45:19.520 --> 00:45:21.520
It's a level that a lot of people are not going to have to care about.
00:45:21.520 --> 00:45:23.540
I think the libraries are the ones that are going to have to care about this
00:45:23.570 --> 00:45:25.520
and who are going to do a lot of this for you.
00:45:25.620 --> 00:45:29.400
Let me throw this out on our way out the door to get to Reuven's topic.
00:45:29.920 --> 00:45:32.860
I would love to see it solidify around async and await.
00:45:33.150 --> 00:45:36.140
And you just await a thing, maybe put a decorator on something.
00:45:36.260 --> 00:45:42.040
say this, this one, I want this to be threaded. I want this to be IO. I want this. And you don't,
00:45:42.100 --> 00:45:46.080
you just use async and await and don't have to think about it, but that's, that's my dream.
00:45:46.800 --> 00:45:47.800
Reuven, what's your dream?
00:45:48.700 --> 00:45:49.740
Wow. How long do you have?
00:45:51.000 --> 00:45:51.600
No, what's your topic?
00:45:52.640 --> 00:45:58.480
So I want to talk about Python ecosystem and funding. When I talk to people with Python
00:45:59.060 --> 00:46:01.820
and I talk to them about it, how it's open source, they're like, oh, right, it's open source. That
00:46:01.940 --> 00:46:05.360
means I can download it for free. And from their perspective, that's sort of where it starts
00:46:05.640 --> 00:46:10.460
and ends. And the notion that people work on it, the notion that needs funding, the notion that
00:46:10.620 --> 00:46:15.060
there's a Python software foundation that supports a lot of these activities, the infrastructure
00:46:15.580 --> 00:46:21.400
is completely unknown to them and even quite shocking for them to hear. But Python is in many
00:46:21.560 --> 00:46:26.720
ways, I think, starting to become a victim of its own success, that we've been dependent on
00:46:27.280 --> 00:46:32.740
companies for a number of years to support developers and development. And we've been
00:46:32.700 --> 00:46:38.080
assuming that the PSF, which gives money out to lots of organizations to run conferences and
00:46:38.290 --> 00:46:43.380
workshops and so forth, can sort of keep scaling up and that they will have enough funding. And
00:46:43.520 --> 00:46:48.500
we've seen a few sort of shocks that system in the last year. Most recently, the PSF announced that
00:46:48.500 --> 00:46:52.680
it was no longer going to be doing versus sort of pared down about a year ago, what it would give
00:46:52.840 --> 00:46:56.620
money for. And then about five months ago, six months ago, I think it was in July or August,
00:46:56.800 --> 00:46:59.700
they said, actually, we're not going to be able to fund anything for about a year now.
00:47:00.000 --> 00:47:17.200
And then there was the government grant, I think from the NSF that they turned down. And I'm not disputing the reasons for that at all. It basically, it said, well, we'll give you the money if you don't worry about diversity and inclusion. And given that that's like a core part of what the PSF is supposed to do, they could not do that without shutting the doors, which would be kind of counterproductive.
00:47:17.540 --> 00:47:26.480
And so I feel like we're not yet there, but we're sort of approaching this, I'm going to term like a problem crisis in funding Python.
00:47:27.040 --> 00:47:32.020
The needs of the community keep growing and growing, whether it's workshops, whether it's PyPI, whether it's conferences.
00:47:32.740 --> 00:47:34.640
And companies are getting squeezed.
00:47:35.160 --> 00:47:41.760
And the number of people, it always shocks me every time there are PSF elections, the incredibly small number of people who vote.
00:47:42.080 --> 00:47:45.680
Which means that, let's assume half the people who are members, third of the people.
00:47:46.000 --> 00:47:50.580
Like for the millions and millions of people who program Python out there, an infinitesimally
00:47:50.650 --> 00:47:53.240
small proportion of them actually join and help to fund it.
00:47:53.430 --> 00:47:56.620
So I'm not quite sure what to do with this other than express concern.
00:47:57.110 --> 00:48:01.760
But I feel like we've got to figure out ways to fund Python and the PSF in new ways that
00:48:01.760 --> 00:48:04.440
will allow it to grow and scale as needed.
00:48:04.620 --> 00:48:05.620
I couldn't agree more.
00:48:06.220 --> 00:48:11.760
Obviously, the PSF is close to my heart because I was on the board for, I think, a total of
00:48:11.920 --> 00:48:15.520
six or nine years or something over, you know, the last 25.
00:48:15.640 --> 00:48:21.520
I was also for six months, I was the interim general manager because Eva left and we hadn't
00:48:21.550 --> 00:48:23.480
hired Deb yet while I was on the board.
00:48:23.670 --> 00:48:28.940
I remember signing the sponsor contracts for the companies that came in wanting to sponsor
00:48:29.170 --> 00:48:29.320
Python.
00:48:29.760 --> 00:48:35.260
And it is like, it's ridiculous how, and I can say this working for a company that is
00:48:35.310 --> 00:48:38.360
one of the biggest sponsors of the PSF and has done so for years.
00:48:38.920 --> 00:48:45.160
It's ridiculous how small those sponsorships are and yet how grateful we were that they
00:48:45.080 --> 00:48:47.860
came in because every single one has such a big impact.
00:48:48.060 --> 00:48:51.620
You can do so much good with the money that comes in.
00:48:52.020 --> 00:48:55.140
We need more corporate sponsorships more than we need.
00:48:55.300 --> 00:49:00.580
Like, I mean, obviously a million people giving us a couple of bucks, giving the PSF, let's
00:49:00.580 --> 00:49:00.900
be clear.
00:49:00.960 --> 00:49:01.780
I'm not on the board anymore.
00:49:01.980 --> 00:49:05.160
Giving the PSF a couple of bucks would be fantastic.
00:49:05.680 --> 00:49:11.880
But I think the big players in the big corporate players where all the AI money is, for instance,
00:49:12.060 --> 00:49:17.380
having done basically no sponsorship of the PSF is mind-boggling. It is a textbook
00:49:18.420 --> 00:49:25.220
tragedy of the commons right there, right? They rely entirely on PyPI and PyPI is run entirely
00:49:25.220 --> 00:49:30.820
with community resources, mostly because of very generous and consistent sponsorship,
00:49:31.180 --> 00:49:38.860
basically by Fastly, but also the other sponsors of the PSF. And yet very large players use those
00:49:38.880 --> 00:49:45.640
resources more than anyone else and don't actually contribute. Georgie Kerr, she wrote this fantastic
00:49:45.730 --> 00:49:52.000
blog post saying pretty much this straight after EuroPython. So EuroPython this year was really big
00:49:52.260 --> 00:49:57.240
actually. And she was wandering around looking at the sponsor booths and the usual players were there,
00:49:57.620 --> 00:50:03.980
but none of these AI companies were there. And the relationship actually between AI, if you want to
00:50:03.800 --> 00:50:09.320
call it that. Let's call it ML and neural networks. And like some of the really big companies and
00:50:09.610 --> 00:50:14.700
Python actually is really complex. Obviously, a lot of these companies and some of us are here,
00:50:15.160 --> 00:50:20.920
employ people to work on Python. Companies like Meta and Google have contributed massively to
00:50:21.140 --> 00:50:27.140
frameworks like PyTorch, TensorFlow, Keras. So it's not as simple a picture as saying cough up money
00:50:27.570 --> 00:50:32.620
all the time. Like there's a more complex picture here, but definitely there are some notable
00:50:32.900 --> 00:50:39.200
absences. And we talked about the volume of money going through. I totally agree with the sentiment.
00:50:39.460 --> 00:50:45.720
When the shortfall came and the grants program had to shut down, we were brainstorming at JetBrains,
00:50:45.820 --> 00:50:51.140
like maybe we can do some sort of, I don't know, donate some more money and call other companies
00:50:51.260 --> 00:50:55.820
to do it. Or we can call on people in the community. And I was like, I don't want to call
00:50:55.820 --> 00:50:59.760
on people in the community to do it because they're probably the same people who are also
00:50:59.780 --> 00:51:05.620
donating their time for Python. Like it's just squeezing people who give so much of themselves
00:51:05.880 --> 00:51:10.700
to this community even more. And it's not sustainable. Like Reuben said, if we keep doing
00:51:10.980 --> 00:51:15.800
this, the whole community is going to collapse. Like I'm sure we've all had our own forms of
00:51:16.030 --> 00:51:19.780
burnout from giving too much. I'm going to pat ourselves on the back here. Everyone on this
00:51:19.940 --> 00:51:25.440
call who works at a company are all sponsors of the PSF. Thank goodness. But there's obviously a
00:51:25.440 --> 00:51:29.100
lot of people not on this call who are not sponsors. And I know personally, I wished every
00:51:29.120 --> 00:51:35.220
company that generated revenue from python usage donated to the psf like and it doesn't see and i
00:51:35.300 --> 00:51:38.440
think part of the problem is some people think it has to be like a hundred thousand dollars it does
00:51:38.520 --> 00:51:43.380
not have to be a hundred thousand dollars now if you can afford that amount please do or more there
00:51:43.430 --> 00:51:48.140
are many ways to donate more than the maximum amount for getting on the website but it's one
00:51:48.140 --> 00:51:51.500
of these funny things where a lot of people just like oh it's not
me right like even startups don't
00:51:51.780 --> 00:51:56.300
some do to give those ones credit but others don't because like oh we're we're burning through
00:51:56.320 --> 00:52:01.840
capital level i was like yeah but we're not we're asking for like less so you'd pay a dev right by
00:52:01.840 --> 00:52:07.280
a lot per year right like the amount we actually asked for to get to the highest tier is still less
00:52:07.340 --> 00:52:12.920
than a common developer in silicon valley if we're gonna price point to a geograph geogra geographical
00:52:13.080 --> 00:52:18.520
location we call kind of comprehend i'm gonna steal a Ned Batchelder's observation here and yeah what
00:52:18.520 --> 00:52:25.740
the psf would be happy with is less than a medium company spends on the tips of expensed meals every
00:52:25.760 --> 00:52:31.340
year. Yeah. Yeah. And it's a long running problem, right? Like, I mean, I've been on the PSF for a
00:52:31.370 --> 00:52:36.380
long time, too. I've not served as many years as Thomas on the board, but I was like executive
00:52:36.470 --> 00:52:40.780
vice president because we had to have someone with that title at some point. It's always been a
00:52:40.940 --> 00:52:45.520
struggle, right? Like I and I also want to be clear, I'm totally appreciative of where we have
00:52:45.630 --> 00:52:50.660
gotten to, right? Because for the longest time, I was just dying for paid staff on the core team.
00:52:50.830 --> 00:52:55.120
And now we have three developers as residents. Thank goodness. Still not enough to be clear.
00:52:55.320 --> 00:52:55.920
I want five.
00:52:56.300 --> 00:52:57.080
And I've always said that,
00:52:57.340 --> 00:52:58.320
but I'll happily take three.
00:52:58.740 --> 00:52:59.500
But it's one of these things
00:52:59.560 --> 00:53:00.480
where it's a constant struggle.
00:53:00.740 --> 00:53:02.360
And it got a little bit better
00:53:02.600 --> 00:53:03.240
before the pandemic
00:53:03.460 --> 00:53:04.480
just because everyone
00:53:04.520 --> 00:53:05.480
was spending on conferences
00:53:05.840 --> 00:53:07.240
and PyCon US is a big driver
00:53:07.320 --> 00:53:08.420
for the Python Software Foundation.
00:53:08.860 --> 00:53:10.300
And I know your Python's a driver
00:53:10.340 --> 00:53:11.560
for the European Society.
00:53:12.140 --> 00:53:12.880
But then COVID hit
00:53:13.340 --> 00:53:14.980
and conferences haven't picked back up.
00:53:15.180 --> 00:53:16.480
And then there's a whole new cohort
00:53:16.880 --> 00:53:19.400
of companies that have come in post-pandemic
00:53:19.740 --> 00:53:20.720
that have never had that experience
00:53:21.020 --> 00:53:22.340
of going to PyCon and sponsoring PyCon.
00:53:22.440 --> 00:53:23.220
And so they don't think about,
00:53:23.360 --> 00:53:25.280
I think, sponsoring PyCon
00:53:25.300 --> 00:53:29.140
the PSF because that's also a big kind of in your face, you should help sponsor this.
00:53:29.570 --> 00:53:33.840
And I think it's led to this kind of lull where offered spending has gone down, new entrants
00:53:33.890 --> 00:53:37.480
into the community have not had that experience and thought about it. And it's led to this kind
00:53:37.480 --> 00:53:43.600
of dearth where, yeah, that PSF had to stop giving out grant money. And it sucks. And I would love
00:53:43.600 --> 00:53:47.600
to see it not be that problem. I want to add one interesting data point that I discovered in
00:53:47.690 --> 00:53:53.800
short. Keep it short. Yes. NumFocus has about twice the budget of the PSF. I was shocked to
00:53:53.820 --> 00:54:00.360
discover this. So basically it is possible to get money from companies to sponsor development of
00:54:00.540 --> 00:54:05.480
Python related projects. And I don't know what they're doing that we aren't. And I think it's
00:54:05.540 --> 00:54:10.960
worth talking and figuring it out. We need a fundraiser and marketer in residence, maybe. Who
00:54:10.960 --> 00:54:17.920
knows? Lauren does a great job, to be clear. The PSF has Lauren and Lauren is that. But it's still
00:54:18.040 --> 00:54:21.980
hard. We have someone doing it full time at the PSF and it's just hard to get companies to give
00:54:22.020 --> 00:54:22.660
cash up cash.
00:54:23.500 --> 00:54:24.880
Yeah, and what do we get in return?
00:54:25.280 --> 00:54:26.080
Well, we already get that.
00:54:26.260 --> 00:54:27.380
So, yeah, I know.
00:54:27.840 --> 00:54:28.580
All right, Barry.
00:54:28.800 --> 00:54:30.220
To just, you know, shift gears
00:54:30.600 --> 00:54:31.620
into a different area,
00:54:32.020 --> 00:54:34.560
something that I've been thinking a lot
00:54:34.720 --> 00:54:36.580
over this past year on the steering council.
00:54:36.860 --> 00:54:38.000
Thomas, I'm sure, is going to be,
00:54:38.160 --> 00:54:39.200
you know, very well aware,
00:54:39.680 --> 00:54:41.620
having been instrumental
00:54:42.040 --> 00:54:45.180
in the lazy imports PEP 810.
00:54:45.620 --> 00:54:47.540
We have to sort of rethink
00:54:48.260 --> 00:54:50.220
how we evolve Python
00:54:50.900 --> 00:54:53.420
and how we pose changes to Python
00:54:53.960 --> 00:54:56.440
and how we discuss those changes in the community.
00:54:56.590 --> 00:54:59.940
Because I think one of the things that I have heard
00:55:00.230 --> 00:55:03.780
over and over and over again is that authoring PEPs
00:55:04.260 --> 00:55:08.780
is incredibly difficult and emotionally draining
00:55:09.500 --> 00:55:10.940
and it's a time sink.
00:55:11.620 --> 00:55:15.860
And leading those discussions on discuss.python.org,
00:55:15.990 --> 00:55:17.520
which we typically call DPO,
00:55:18.420 --> 00:55:21.400
can be toxic at times and very difficult.
00:55:21.800 --> 00:55:24.340
So one of the things that I realized
00:55:24.650 --> 00:55:25.960
as I was thinking about this
00:55:25.960 --> 00:55:28.760
is that peps are 25 years old now, right?
00:55:29.240 --> 00:55:30.820
So we've had this,
00:55:31.320 --> 00:55:32.980
and
not only just peps are old,
00:55:33.110 --> 00:55:35.700
but like we've gone through at least two,
00:55:35.890 --> 00:55:38.520
if not more sort of complete revolutions
00:55:38.590 --> 00:55:40.040
in the way we discuss things.
00:55:40.170 --> 00:55:42.900
You know, the community has grown incredibly.
00:55:43.290 --> 00:55:45.720
The developer community is somewhat larger,
00:55:46.180 --> 00:55:47.660
but just the number of people
00:55:47.680 --> 00:55:54.380
who are using Python and who have an interest in it has grown exponentially. So it has become
00:55:54.900 --> 00:56:01.400
really difficult to evolve the language in the standard library and the interpreter. And we need
00:56:01.400 --> 00:56:08.760
to sort of think about how we can make this easier for people and not lose the voice of the user.
00:56:09.160 --> 00:56:14.500
And the number of people who actually engage in topics on DPO is the tip of the iceberg. You know,
00:56:14.600 --> 00:56:24.740
We've got millions and millions of users out there in the world who, for example, lazy imports will affect, free threading will affect and don't even know that they have a voice.
00:56:24.890 --> 00:56:31.200
And maybe we have to basically represent that, but we have to do it in a much more collaborative and positive way.
00:56:31.740 --> 00:56:33.400
That's something that I've been thinking about a lot.
00:56:33.500 --> 00:56:43.580
And whether or not I'm on the steering council next year, I think this is something that I'm going to spend some time on trying to think about, you know, talk to people about ways we can make this easier for everyone.
00:56:43.820 --> 00:56:47.340
The diversity of use cases for Python in the last couple of years.
00:56:47.390 --> 00:56:48.000
So complex.
00:56:48.420 --> 00:56:48.980
Yes, exactly.
00:56:49.210 --> 00:56:53.060
It should also be preface that Barry created the PEP process. He should have started that one.
00:56:55.780 --> 00:56:57.280
It is that old.
00:56:57.610 --> 00:56:57.720
Yeah.
00:56:58.540 --> 00:57:02.120
By the way, just so everyone knows, these are not ages jokes to be mean to Barry.
00:57:02.340 --> 00:57:07.080
We've always known Barry long enough that we know Barry's okay with us making these jokes.
00:57:07.110 --> 00:57:07.900
To be very, very clear.
00:57:07.900 --> 00:57:11.920
Also, I am almost as old as Barry, although I don't look as old as Barry.
00:57:12.780 --> 00:57:14.300
Yeah, we're all over from the same age anyways.
00:57:15.300 --> 00:57:18.600
Yeah, Barry and I have known each other for 25 years,
00:57:18.960 --> 00:57:21.760
and I've always made these jokes of him.
00:57:21.940 --> 00:57:26.220
So it is different when you know each other in person.
00:57:26.420 --> 00:57:27.000
Let's put it that way.
00:57:28.840 --> 00:57:30.900
For the PEP process, I think for a lot of people,
00:57:31.160 --> 00:57:35.080
it's not obvious how difficult the process is.
00:57:35.260 --> 00:57:36.760
I mean, it wasn't even obvious to me.
00:57:37.260 --> 00:57:40.720
I saw people avoiding writing peps multiple times,
00:57:40.820 --> 00:57:42.940
and I was upset, like on the steering council, right?
00:57:43.380 --> 00:57:45.280
I saw people making changes where I thought,
00:57:45.440 --> 00:57:46.820
this is definitely something
00:57:47.020 --> 00:57:48.420
that should have been discussed in a PEP
00:57:48.700 --> 00:57:51.220
and the discussion should be recorded in a PEP and all that.
00:57:51.540 --> 00:57:54.600
And I didn't understand why they didn't until,
00:57:55.260 --> 00:57:56.800
basically until PEP 810.
00:57:56.980 --> 00:57:58.920
So I did PEP 779,
00:57:58.990 --> 00:58:02.400
which was the giving free threading supported status
00:58:02.820 --> 00:58:03.660
at the start of the year.
00:58:03.980 --> 00:58:06.200
And the discussion there was, you know,
00:58:06.560 --> 00:58:08.540
sort of as expected and it's already,
00:58:08.760 --> 00:58:10.280
was already an accepted PEP.
00:58:10.600 --> 00:58:12.560
It was just the question of how does it become supported?
00:58:12.990 --> 00:58:14.240
That one wasn't too exhausting.
00:58:14.770 --> 00:58:20.040
And then we got to Lazy Imports, which was Pablo, who is another steering council member,
00:58:20.460 --> 00:58:24.720
as well as a bunch of other contributors, including me and two of my co-workers and
00:58:24.820 --> 00:58:28.900
one of my former co-workers, who had all had a lot of experience with Lazy Imports, but
00:58:29.040 --> 00:58:31.540
not necessarily as much experience with the PEP process.
00:58:32.140 --> 00:58:37.680
And Pablo took the front seat because he knew the PEP process and he's done like five PEPs
00:58:37.680 --> 00:58:40.000
in the last year or something, some ridiculous number.
00:58:40.360 --> 00:58:48.880
And he shared with us the vitriol he got for like offline for the, just the audacity of proposing
00:58:49.120 --> 00:58:54.240
something that people disagreed with or something. And that was like, this is a technical suggestion.
00:58:54.300 --> 00:58:59.740
This is not a code of conduct issue where I have received my fair share of vitriol around.
00:59:00.160 --> 00:59:06.200
This is a technical discussion. And yet he gets this, these ridiculous accusations in his mailbox.
00:59:06.740 --> 00:59:11.980
And for some reason, only the primary author gets it as well, which is just weird to me.
00:59:12.140 --> 00:59:13.360
But people are lazy.
00:59:13.780 --> 00:59:15.420
Thomas is what I think you just said.
00:59:15.640 --> 00:59:24.500
Remember, the steering council exists because Guido was the got the brunt of this for Pet 572, which was the walrus operator.
00:59:24.680 --> 00:59:30.660
Right. Which is just like this minor little syntactic thing that is kind of cool when you need it.
00:59:31.020 --> 00:59:42.340
But like just the amount of anger and negative energy and vitriol that he got over that was enough to for him to just say, I'm out, you know, and you
guys figure it out.
00:59:42.400 --> 00:59:48.280
And that cannot be an acceptable way to discuss the evolution of the language.
00:59:48.420 --> 00:59:55.100
Especially since apparently now every single PEP author of any contentious or semi contentious PEP.
00:59:55.420 --> 00:59:58.820
Although I have to say, Pep 810 had such broad support.
00:59:59.030 --> 01:00:00.940
It was hard to call it contentious.
01:00:01.060 --> 01:00:04.240
It's just there's a couple of very loud opinions, I guess.
01:00:04.600 --> 01:00:06.720
And I'm not saying we shouldn't listen to people.
01:00:06.770 --> 01:00:10.980
We should definitely listen to especially contrary opinions.
01:00:11.130 --> 01:00:12.460
But there has to be a limit.
01:00:12.750 --> 01:00:15.880
There has to be an acceptable way of bringing things up.
01:00:15.890 --> 01:00:20.760
There has to be an acceptable way of saying, hey, you didn't actually read the Pep.
01:00:21.120 --> 01:00:26.720
please go back and reconsider everything you said after you fully digested the things,
01:00:27.180 --> 01:00:29.020
because everything's already been addressed in the PEP.
01:00:29.400 --> 01:00:37.480
It's just really hard to do this in a way that doesn't destroy the relationship with the person you're telling this, right?
01:00:37.980 --> 01:00:44.020
It's hard to tell people, hey, I'm not going to listen to you because you haven't, you know, you've done a bad job.
01:00:44.520 --> 01:00:45.860
You've chosen not to inform yourself.
01:00:46.220 --> 01:00:54.860
I think you make another really strong point, Thomas, which is that there have been changes that have been made to Python that really should have been a PEP.
01:00:55.220 --> 01:01:01.480
And they aren't because people don't want to go through core developers, don't want to go through this gauntlet.
01:01:01.570 --> 01:01:03.700
And so they'll create a PR and then that.
01:01:03.700 --> 01:01:06.880
But that's also not good because then, you know, we don't have that.
01:01:06.970 --> 01:01:10.520
We don't have the right level of consideration.
01:01:11.160 --> 01:01:19.880
And you think about the way that, you know, if you're in your job and you're making a change to something in your job, you have a very close relationship to your teammates.
01:01:20.220 --> 01:01:26.000
And so you have that kind of respect and hopefully, right, like compassion and consideration.
01:01:26.720 --> 01:01:35.600
And you can have a very productive discussion about a thing and you may win some arguments and you may lose some arguments, but the team moves forward as one.
01:01:35.840 --> 01:01:38.500
And I think we've lost a bit of that in Python.
01:01:38.920 --> 01:01:39.940
So that's not great.
01:01:40.000 --> 01:01:50.220
I think society in general could use a little more civility and kindness, especially to strangers that they haven't met in forums, social media, driving, you name it.
01:01:50.760 --> 01:01:54.160
Okay, but we're not going to solve that here,
I'm sure.
01:01:54.420 --> 01:01:56.920
So instead, let's do Gregory's topic.
01:01:57.260 --> 01:02:04.160
Hey, I'm going to change topics quite a bit, but I wanted to call 2025 the year of type checking and language server protocols.
01:02:04.800 --> 01:02:13.780
So many of us probably have used tools like mypy to check to see if the types line up in our code or whether or not we happen to be overriding functions correctly.
01:02:14.170 --> 01:02:19.980
And so I've used mypy for many years and loved the tool and had a great opportunity to chat with the creator of it.
01:02:20.260 --> 01:02:23.800
And I integrate that into my CI and it's really been wonderful.
01:02:24.180 --> 01:02:28.600
And I've also been using a lot of LSPs, like, for example, Pyright or PyLands.
01:02:28.900 --> 01:02:33.860
But in this year, one of the things that we've seen is, number one, Pyrefly from the team at Meta.
01:02:34.160 --> 01:02:36.600
We've also seen ty from the team at Astral.
01:02:36.930 --> 01:02:38.600
And there's another one called Zuban.
01:02:38.910 --> 01:02:40.760
And Zuban is from David Halter.
01:02:41.120 --> 01:02:43.480
David was also the person who created JEDI,
01:02:43.770 --> 01:02:47.620
which is another system in Python that helped with a lot of LSP tasks.
01:02:48.040 --> 01:02:51.440
What's interesting about all three of the tools that I just mentioned
01:02:51.800 --> 01:02:53.320
is that they're implemented in Rust,
01:02:53.800 --> 01:02:58.080
and they have taken a lot of the opportunity to make the type checker
01:02:58.260 --> 01:03:01.040
and or the LSP significantly faster.
01:03:01.480 --> 01:03:07.640
So for me, this has changed how I use the LSP or the type checker and how frequently I use it.
01:03:07.880 --> 01:03:16.580
And in my experience, it has helped me to take things that might take tens of seconds or hundreds of seconds and cut them down often to less than a second.
01:03:17.080 --> 01:03:23.860
And it's really changed the way in which I'm using a lot of the tools like ty or Pyrefly or Zuban.
01:03:24.180 --> 01:03:31.800
So I can have some more details if I'm allowed to share, Michael, but I would say 2025 is the year of type checkers and LSPs.
01:03:31.980 --> 01:03:34.320
I think given the timing, let's have people give some feedback.
01:03:34.540 --> 01:03:38.680
I personally have been using Pyrefly a ton and am a big fan of it.
01:03:38.740 --> 01:03:42.600
I don't know if I'm allowed to have an opinion that isn't Pyrefly is awesome.
01:03:43.380 --> 01:03:48.500
I mean, I'm not on the Pyrefly team, but I do regularly chat with people from the Pyrefly team.
01:03:49.040 --> 01:03:50.840
Tell people real quick what it is, Thomas.
01:03:51.180 --> 01:03:55.740
So Pyrefly is Meta's attempt at a Rust-based type checker.
01:03:56.140 --> 01:03:57.920
And so it's very similar to ty.
01:03:58.360 --> 01:04:01.240
Started basically at the same time, a little later.
01:04:01.920 --> 01:04:06.100
Meta originally had a
type checker called Pyre, which was written in OCaml.
01:04:06.480 --> 01:04:09.020
They basically decided to start a rewrite in Rust.
01:04:09.300 --> 01:04:11.220
And then that really took off.
01:04:11.380 --> 01:04:13.660
And that's where we're going now.
01:04:13.900 --> 01:04:14.000
Yeah.
01:04:14.140 --> 01:04:14.220
Yeah.
01:04:14.230 --> 01:04:17.960
I don't know what I can say because I'm actually on the same team as the Pylance team.
01:04:18.460 --> 01:04:20.780
So, but no, I mean, I think it's good.
01:04:21.060 --> 01:04:22.920
I think this is one of those interesting scenarios
01:04:23.660 --> 01:04:25.180
where some people realize like,
01:04:25.340 --> 01:04:27.260
you know what, we're going to pay the penalty
01:04:28.059 --> 01:04:30.580
of writing a tool in a way that's faster,
01:04:30.740 --> 01:04:31.640
but makes us go slower
01:04:31.880 --> 01:04:33.220
because the overall win for the community
01:04:33.420 --> 01:04:34.840
is going to be a good win.
01:04:34.900 --> 01:04:36.460
So it's worth that headache, right?
01:04:36.780 --> 01:04:38.240
Not to say I don't want to scare people off
01:04:38.360 --> 01:04:39.640
from writing Rust, but let's be honest,
01:04:39.700 --> 01:04:41.040
it takes more work to write Rust code
01:04:41.140 --> 01:04:42.360
than it does take to write Python code.
01:04:42.940 --> 01:04:44.700
But some people chose to make that trade off
01:04:44.740 --> 01:04:45.960
and we're all benefiting from it.
01:04:46.260 --> 01:04:47.180
The one thing I will say
01:04:47.280 --> 01:04:48.280
that's kind of interesting from this
01:04:48.620 --> 01:04:50.060
that hasn't gotten a lot of play yet
01:04:50.120 --> 01:04:50.940
because it's still being developed,
01:04:51.060 --> 01:05:05.580
But Pylance is actually working with the Pyrefly team to define a type server protocol, TSP, so that a lot of these type servers can just kind of feed the type information to a higher level LSP and let that LSP handle the stuff like symbol renaming and all that stuff.
01:05:05.620 --> 01:05:12.300
Right. Because the key thing here and the reason there's so many different type checkers is there are there is a spec.
01:05:12.460 --> 01:05:16.500
Right. And everyone's trying to implement it. But there's differences like in type in terms of type inferencing.
01:05:16.800 --> 01:05:19.100
And if I actually go listen to Michael's interview,
01:05:19.780 --> 01:05:21.460
talk Python to me with the Pyrefly team,
01:05:21.560 --> 01:05:23.080
they actually did a nice little explanation
01:05:23.140 --> 01:05:24.760
of the difference between Pyright's approach
01:05:25.600 --> 01:05:26.860
and Pyrefly's approach.
01:05:27.420 --> 01:05:28.360
And so there's a bit of variance.
01:05:28.640 --> 01:05:31.140
But for instance, I think there's some talk now
01:05:31.360 --> 01:05:32.440
of trying to like, how do we make it
01:05:32.440 --> 01:05:33.520
so everyone doesn't have to reimplement
01:05:33.640 --> 01:05:34.980
how to rename a symbol, right?
01:05:35.100 --> 01:05:35.680
That's kind of boring.
01:05:35.840 --> 01:05:37.300
That's not where the interesting work is.
01:05:37.640 --> 01:05:39.940
And that's not performant from perspective of
01:05:40.180 --> 01:05:42.920
you want instantaneously to get that squiggly red line
01:05:43.500 --> 01:05:45.039
in whether it's VS Code
01:05:45.040 --> 01:05:48.220
or it's in PyCharm or whatever your editor is, right?
01:05:48.380 --> 01:05:50.340
You want to get it as fast as possible,
01:05:50.500 --> 01:05:51.340
but the rename-
01:05:51.660 --> 01:05:51.860
Jupyter.
01:05:52.160 --> 01:05:52.280
Jupyter.
01:05:52.400 --> 01:05:53.520
No, not Emacs.
01:05:53.520 --> 01:05:53.860
Everything but Emacs.
01:05:53.900 --> 01:05:56.020
No, not Emacs.
01:05:56.020 --> 01:05:57.640
Just to bring things full circle,
01:05:57.640 --> 01:06:00.500
it's that focus on user experience, right?
01:06:00.680 --> 01:06:03.020
Which is, yes, you want that squiggly line,
01:06:03.020 --> 01:06:04.740
but when things go wrong,
01:06:04.740 --> 01:06:05.640
when your type checker says,
01:06:05.640 --> 01:06:07.780
oh, you've got a problem,
01:06:07.780 --> 01:06:11.820
you know, like I think about as an analogy,
01:06:11.820 --> 01:06:14.600
how Pablo has done an amazing amount of work
01:06:14.860 --> 01:06:14.920
on the error reporting, right?
01:06:15.020 --> 01:06:26.460
When you get an exception and, you know, now you have a lot more clues about what is it that I actually have to change to make the tool, you know, to fix the problem, right?
01:06:26.780 --> 01:06:39.740
Like so many times years ago, you know, when people were using mypy, for example, and they'd have some complex failure of their type annotations and have absolutely no idea what to do about it.
01:06:40.140 --> 01:06:44.480
And so getting to a place where now we're not just telling people you've done it wrong,
01:06:44.960 --> 01:06:48.560
but also here's some ideas about how to fix it.
01:06:49.540 --> 01:06:54.340
I think this is a full circle here because honestly, using typing in your Python code
01:06:54.600 --> 01:06:57.020
gives a lot of context to the AI when you ask for help.
01:06:57.200 --> 01:06:59.320
If you just give it a fragment and it can't work with it.
01:06:59.480 --> 01:07:00.080
That's true.
01:07:00.360 --> 01:07:06.320
And also, if you can teach your AI agent to use the type checkers and use the LSPs,
01:07:06.640 --> 01:07:08.660
it will also generate better code for you.
01:07:09.000 --> 01:07:24.320
I think the one challenge I would add to what Barry said a moment ago is that if you're a developer and you're using, say, three or four type checkers at the same time, you also have to be careful about the fact that some of them won't flag an error that the other one will flag.
01:07:24.680 --> 01:07:36.300
So I've recently written Python programs and even built a tool with one of my students named Benedek that will automatically generate Python
programs that will cause type checkers to disagree with each other.
01:07:39.320 --> 01:07:44.580
I will flag it as an error, but none of the other tools will flag it as an error.
01:07:45.020 --> 01:07:50.020
And there are also cases where the new tools will all agree with each other, but disagree with mypy.
01:07:50.360 --> 01:07:53.100
So there is a type checker conformance test suite.
01:07:53.280 --> 01:07:57.280
But I think as developers, even though it might be the year of LSP and type checker,
01:07:57.300 --> 01:08:02.980
we also have to be aware of the fact that these tools are maturing and there's still disagreement among them.
01:08:03.400 --> 01:08:07.700
and also just different philosophies when it comes to how to type check and how to infer.
01:08:08.240 --> 01:08:12.320
And so we have to think about all of those things as these tools mature and become part of our ecosystem.
01:08:12.480 --> 01:08:14.040
Yeah, Greg, that last point is important.
01:08:14.240 --> 01:08:17.940
Out of curiosity, how did the things where the type checkers disagree
01:08:18.560 --> 01:08:21.380
match up with the actual runtime behavior of Python?
01:08:21.980 --> 01:08:24.480
Was it like false positives or false negatives?
01:08:24.880 --> 01:08:25.759
That's a really good question.
01:08:26.000 --> 01:08:29.900
I'll give you more details in the show notes because we actually have it in a GitHub repository
01:08:30.220 --> 01:08:31.339
and I can share it with people.
01:08:31.740 --> 01:08:43.200
But I think some of it might simply be related to cases where mypy is more conformant to the spec, but the other new tools are not as conformant.
01:08:43.370 --> 01:08:48.180
So you can import overload from typing and then have a very overloaded function.
01:08:48.720 --> 01:08:59.960
And mypy will actually flag the fact that it's an overloaded function with multiple signatures, whereas Pyright and Pyrefly and Zuban will not actually flag that, even though they should.
01:09:00.160 --> 01:09:02.819
Another big area is optional versus not optional.
01:09:03.380 --> 01:09:03.480
Yes.
01:09:03.700 --> 01:09:08.960
Like, are you allowed to pass a thing that is an optional string when the thing accepts a string?
01:09:09.180 --> 01:09:10.380
Some stuff's like, yeah, it's probably fine.
01:09:10.420 --> 01:09:11.359
Others are like, no, no, no.
01:09:11.400 --> 01:09:13.540
This is an error that you have to do a check.
01:09:13.720 --> 01:09:23.259
And if you want to switch type checkers, you might end up with a thousand warnings that you didn't previously had because of an intentional difference of opinion on how strict to be, I think.
01:09:23.400 --> 01:09:23.560
Yeah.
01:09:23.680 --> 01:09:29.220
So you have to think about false positives and false negatives when you're willing to break the build because of a type error.
01:09:29.299 --> 01:09:31.720
All of those things are things you have to factor in.
01:09:31.990 --> 01:09:37.040
But to go quickly to this connection to AI, I know it's only recently, but the Pyrefly
01:09:37.130 --> 01:09:41.859
team actually announced that they're making Pyrefly work directly with Pydantic AI.
01:09:42.400 --> 01:09:46.640
So there's going to be an interoperability between those tools so that when you're building
01:09:46.710 --> 01:09:52.160
an AI agent using Pydantic AI, you can also then have better guarantees when you're using
01:09:52.339 --> 01:09:53.580
Pyrefly as your type checker.
01:09:53.680 --> 01:09:57.659
It makes total sense, though, because then the reasoning LLM that's at the core of the
01:09:57.560 --> 01:10:04.220
agent can actually have that information before it tries to execute the code and you don't get in that
01:10:04.220 --> 01:10:09.840
loop that they often get in. You can correct it before it runs. Yeah, really good point. I want to
01:10:09.840 --> 01:10:14.660
just sort of express my appreciation to all the people working on this typing stuff. As someone
01:10:14.820 --> 01:10:27.520
who's come from many, many years in dynamic languages, I was always like, oh, typing. Those
01:10:27.540 --> 01:10:32.740
E, I love seeing how easy it is for people to ease into it when they're in Python.
01:10:32.980 --> 01:10:33.900
It's not all or nothing.
01:10:34.460 --> 01:10:36.660
C, I love the huge number of tools.
01:10:36.760 --> 01:10:39.160
The competition in this space is really exciting.
01:10:39.640 --> 01:10:40.620
And D, guess what?
01:10:40.840 --> 01:10:41.960
It really, really does help.
01:10:42.240 --> 01:10:46.760
And I'll even add an E, which is my students who come from Java, C++, C#, and so forth
01:10:47.140 --> 01:10:47.920
feel relief.
01:10:48.400 --> 01:10:53.080
They find that without type checking, it's like doing a trapeze act without a safety
01:10:53.220 --> 01:10:53.400
net.
01:10:53.760 --> 01:10:57.160
And so they're very happy to have that typing in there,
01:10:57.380 --> 01:10:57.880
typings in there.
01:10:58.200 --> 01:10:59.240
So kudos to everyone.
01:10:59.480 --> 01:11:00.620
All right, folks, we are out of time.
01:11:00.940 --> 01:11:03.040
This could literally go for hours longer.
01:11:04.640 --> 01:11:05.300
It was a big year.
01:11:05.760 --> 01:11:06.720
It was a big year,
01:11:06.860 --> 01:11:09.940
but I think we need to just have a final word.
01:11:10.800 --> 01:11:12.380
I'll start and we'll just go around.
01:11:12.720 --> 01:11:14.200
So my final thought here is,
01:11:14.660 --> 01:11:16.700
we've talked about some things that are negatives
01:11:17.060 --> 01:11:19.100
or sort of downers or whatever here and there,
01:11:19.640 --> 01:11:22.860
but I still think it's an incredibly exciting time
01:11:22.880 --> 01:11:26.080
To be a developer, data scientist, there's so much opportunity out there.
01:11:26.600 --> 01:11:29.480
There's so many things to learn and take advantage of and stay on top of.
01:11:29.640 --> 01:11:30.280
And amazing.
01:11:30.600 --> 01:11:32.860
Every day is
slightly more amazing than the previous day.
01:11:33.040 --> 01:11:33.860
So I love it.
01:11:34.080 --> 01:11:34.900
Gregory, let's go to you next.
01:11:35.000 --> 01:11:35.580
Let's go around the circle.
01:11:35.780 --> 01:11:39.400
Yeah, I wanted to give a shout out to all of the local Python conferences.
01:11:40.200 --> 01:11:43.660
I actually, on a regular basis, have attended the PyOhio conference.
01:11:44.280 --> 01:11:45.320
And it is incredible.
01:11:45.620 --> 01:11:48.460
The organizers do an absolutely amazing job.
01:11:49.000 --> 01:11:54.120
And they have it hosted on a campus, oftentimes at Ohio State or Cleveland State University.
01:11:54.820 --> 01:12:00.740
And incredibly, PyOhio is a free conference that anyone can attend with no registration fee.
01:12:01.050 --> 01:12:06.180
So Michael, on a comment that I think is really positive, wow, I'm so excited about the regional
01:12:06.430 --> 01:12:08.340
Python conferences that I've been able to attend.
01:12:08.500 --> 01:12:08.660
Thomas.
01:12:09.020 --> 01:12:10.640
Wow, I didn't expect this.
01:12:10.840 --> 01:12:16.520
So I think I want to give a shout out to new people joining the community and also joining
01:12:16.540 --> 01:12:22.060
just core developer team as triagers or it's just drive by commenters. I know we harped a little bit
01:12:22.160 --> 01:12:27.520
about people, you know, giving strong opinions and discussions, but I always look to the far future
01:12:27.720 --> 01:12:33.300
as well as the near future. And we always need new people. We need new ideas. We need new opinions. So
01:12:33.840 --> 01:12:38.960
yeah, I'm, I'm excited that there's still people joining and signing up and even when it's
01:12:39.240 --> 01:12:44.140
thankless work. So I guess I want to say thank you to people doing all the thankless work. Jodi.
01:12:44.380 --> 01:12:51.740
Yeah, I want to say this is actually really only my third year or so really in the Python community.
01:12:52.000 --> 01:12:54.220
So before that, I was just sort of on the fringes, right?
01:12:54.460 --> 01:12:58.020
And after I started advocacy, I started going to the conferences and meeting people.
01:12:58.660 --> 01:13:04.680
And I think I didn't kind of get how special the community was until I watched the Python documentary this year.
01:13:04.920 --> 01:13:11.840
And I talked to Paul about this, Paul Everitt afterwards, also made fun of him for his like early 2000s fashion.
01:13:11.940 --> 01:13:18.400
but I think, yeah, like I'm a relative newcomer to this community and you've all made me feel so
01:13:18.600 --> 01:13:24.000
welcome. And I guess I want to thank all the incumbents for everything you've done to make
01:13:24.120 --> 01:13:30.100
this such a special tech community for minorities and everyone, newbies, you know, Python,
01:13:30.610 --> 01:13:32.860
Python is love. Oh, geez. How am I supposed to follow that?
01:13:37.230 --> 01:13:41.900
I think one of the interesting things that we're kind of looping on here is
01:13:41.920 --> 01:13:45.340
I think the language evolution has slowed down, but it's obviously not stopped, right?
01:13:45.520 --> 01:13:48.640
Like as Thomas pointed out, there's a lot more stuff happening behind the scenes.
01:13:49.580 --> 01:13:54.660
Lazy imports are coming, and that was a syntactic change, which apparently brings out the mean side of some people.
01:13:55.100 --> 01:13:57.900
And we've obviously got our challenges and stuff, but things are still going.
01:13:58.180 --> 01:13:59.380
We're still looking along.
01:13:59.490 --> 01:14:04.900
We're still trying to be an open, welcoming place for people like Jody and everyone else who's new coming on over
01:14:05.100 --> 01:14:11.080
and to continue to be a fun place for all of us slightly grain-beard people who have been here for a long time
01:14:11.100 --> 01:14:12.000
to make us want to stick around.
01:14:12.300 --> 01:14:15.140
I think it's just more of the same, honestly.
01:14:15.380 --> 01:14:17.820
It's all of us just continuing to do what we can
01:14:17.840 --> 01:14:20.320
to help out to keep this community being a great place.
01:14:20.520 --> 01:14:22.480
And it all just keeps going forward.
01:14:22.760 --> 01:14:23.860
And I'll just end with,
01:14:23.900 --> 01:14:25.920
if you work for a company that's not sponsored the PSF,
01:14:26.000 --> 01:14:26.540
please do so.
01:14:26.600 --> 01:14:30.680
It's rare to have, I mean, a programming language
01:14:30.900 --> 01:14:31.840
or any sort of tool
01:14:32.680 --> 01:14:36.040
where it is both really, really beneficial to your career
01:14:36.640 --> 01:14:38.860
and you get to hang out with really special,
01:14:39.320 --> 01:14:40.420
nice, interesting people.
01:14:40.980 --> 01:14:44.680
And it's easy to take all that for granted if you've been steeped in the community.
01:14:45.110 --> 01:14:48.520
I went to a conference about six months ago, a non-Python conference.
01:14:49.180 --> 01:14:54.760
And that was shocking to me to discover that all the speakers were from advertisers and sponsors.
01:14:55.400 --> 01:14:57.240
Everything was super commercialized.
01:14:57.580 --> 01:15:00.440
People were not interested in just like hanging out and sharing with each other.
01:15:00.880 --> 01:15:04.980
And it was a shock to me because I've been to basically only Python conferences for so many years.
01:15:05.050 --> 01:15:08.300
I was like, oh, that's not the norm in the industry.
01:15:08.800 --> 01:15:15.640
So we've got something really special going that not only is good for the people, but good for everyone's careers and mutually reinforcing and helping each other.
01:15:16.120 --> 01:15:17.580
And that's really fantastic.
01:15:17.610 --> 01:15:18.480
And we should appreciate that.
01:15:19.140 --> 01:15:20.280
Barry, final word.
01:15:20.520 --> 01:15:25.340
Thomas stole my thunder just a little bit, but just to tie a couple of these ideas together.
01:15:25.960 -->
01:15:29.360
Python, and you know, Brett said this, right?
01:15:29.520 --> 01:15:33.120
This is Python is the community or the community is Python.
01:15:33.580 --> 01:15:38.240
There's no company that is telling anybody what Python should be.
01:15:38.600 --> 01:15:39.980
Python is what we make it.
01:15:40.460 --> 01:15:43.840
And, you know, as folks like myself get a little older
01:15:44.140 --> 01:15:47.500
and, you know, and we have younger people
01:15:47.700 --> 01:15:48.620
coming into the community,
01:15:48.920 --> 01:15:50.760
both developers and everything else
01:15:51.280 --> 01:15:54.320
who are shaping Python into their vision.
01:15:54.700 --> 01:15:56.120
I encourage you,
01:15:56.380 --> 01:15:58.360
if you've thought about becoming a core dev,
01:15:58.720 --> 01:15:59.440
find a mentor.
01:15:59.660 --> 01:16:01.320
There are people out there that will help you.
01:16:01.520 --> 01:16:05.100
If you want to be involved in the community, the PSF,
01:16:05.460 --> 01:16:06.260
you know, reach out.
01:16:06.320 --> 01:16:08.560
There are people who will help guide you
01:16:08.580 --> 01:16:15.680
this community. You can be involved. Do not let any self-imposed limitations stop you from
01:16:16.740 --> 01:16:22.440
becoming part of the Python community in the way that you want to. And eventually run for the
01:16:22.600 --> 01:16:29.320
steering council because we need many, many, many more candidates next year. And you don't need any
01:16:29.580 --> 01:16:33.940
qualifications either because I'm a high school dropout and I never went to college or anything.
01:16:34.640 --> 01:16:35.400
And look at me.
01:16:35.820 --> 01:16:37.600
And I have a PhD and I will tell you,
01:16:37.610 --> 01:16:39.600
I did not need all that to become a Python developer
01:16:39.830 --> 01:16:41.980
because I was the Python developer before I got the PhD.
01:16:42.540 --> 01:16:43.360
I'm a bass player.
01:16:43.430 --> 01:16:44.960
So if I can do it, anybody can do it.
01:16:47.920 --> 01:16:49.120
Thank you everyone for being here.
01:16:49.520 --> 01:16:50.860
This awesome look back in the air
01:16:50.940 --> 01:16:52.220
and I really appreciate you all taking the time.
01:16:52.420 --> 01:16:52.860
Thank you, Michael.
01:16:53.080 --> 01:16:54.440
Thanks everybody.
01:16:54.830 --> 01:16:55.140
Bye everybody.
01:16:57.320 --> 01:16:59.700
This has been another episode of Talk Python To Me.
01:17:00.080 --> 01:17:00.800
Thank you to our sponsors.
01:17:01.010 --> 01:17:02.300
Be sure to check out what they're offering.
01:17:02.500 --> 01:17:03.860
It really helps support the show.
01:17:04.400 --> 01:17:07.200
Look into the future and see bugs before they make it to production.
01:17:08.000 --> 01:17:13.280
Sentry's Seer AI Code Review uses historical error and performance information at Sentry
01:17:13.660 --> 01:17:17.620
to find and flag bugs in your PRs before you even start to review them.
01:17:18.280 --> 01:17:20.100
Stop bugs before they enter your code base.
01:17:20.620 --> 01:17:24.500
Get started at talkpython.fm/seer-code-review.
01:17:25.020 --> 01:17:26.880
If you or your team needs to learn Python,
01:17:27.080 --> 01:17:30.560
we have over 270 hours of beginner and advanced courses
01:17:30.560 --> 01:17:37.160
on topics ranging from complete beginners to async code, Flask, Django, HTMX, and even LLMs.
01:17:37.480 --> 01:17:39.820
Best of all, there's no subscription in sight.
01:17:40.320 --> 01:17:41.980
Browse the catalog at talkpython.fm.
01:17:42.680 --> 01:17:46.060
And if you're not already subscribed to the show on your favorite podcast player,
01:17:46.700 --> 01:17:47.340
what are you waiting for?
01:17:47.980 --> 01:17:49.780
Just search for Python in your podcast player.
01:17:49.880 --> 01:17:50.760
We should be right at the top.
01:17:51.180 --> 01:17:53.940
If you enjoy that geeky rap song, you can download the full track.
01:17:54.160 --> 01:17:56.060
The link is actually in your podcast blur show notes.
01:17:56.860 --> 01:17:58.220
This is your host, Michael Kennedy.
01:17:58.540 --> 01:17:59.680
Thank you so much for listening.
01:17:59.980 --> 01:18:00.620
I really appreciate it.
01:18:01.120 --> 01:18:01.800
I'll see you next time.
01:18:27.680 --> 01:18:29.420
I think is the norm.