#498: A Tiny Episode

00:00 Hello and welcome to Python Bytes, where we deliver Python news and headlines directly to your buds.

00:05 This is unbelievably episode 498 coming up on episode 500,

00:10 recorded September 29th, 2026.

00:13 I'm Michael Kennedy.

00:14 And I'm Calvin Hendrix Parker.

00:15 Check us out on all the socials.

00:18 We put all the links if you want to interact with us there at the top of the show.

00:21 This episode is brought to you by us.

00:23 So be sure to check out Six Feet Up if you have an amazing,

00:26 you would like, People who've done a lot of problem solving in the Python space for many, many years.

00:31 Reach out to Calvin and learn about Python.

00:34 Got some courses over at Talk Python.

00:36 They had new things coming.

00:37 Nothing to announce yet, but a lot of work has been going lately on a new thing over there.

00:43 So I'm very excited.

00:44 Check out the newsletter.

00:45 Just visit the homepage.

00:46 Click newsletter.

00:47 We got lots of cool things to send to you there.

00:49 And with that, you have scary news.

00:52 Are you going to scare us?

00:53 Hopefully not too much.

00:54 It's not quite October, although I do see lots and lots of decorations out in the stores these days for Halloween.

01:01 But there is unfortunately...

01:02 It's like a jump scare, right?

01:03 A little bit, yeah.

01:04 There's been a supply chain attack and there's no shortage of these over the last couple of years,

01:10 probably due to the proliferation of AI and the tools that attackers can use to build these different exploits.

01:15 This one is around MemTensor and MemoryOS.

01:18 So some people may not be affected by this, but I think the real true story here is to follow

01:25 how it's done and to protect your CI pipelines ultimately.

01:29 So on September 23rd, an attacker published a backdoor into MemoryOS.

01:34 So if you're using OpenClaw or if you have installed a MemoryOS

01:38 skill into your agentic harness, you could be susceptible to this.

01:43 So look for these various versions 0.1.21 on the plugin there's or that's the bad versions and 0.1.23 and 25 and then the

01:53 memory os version 0.2 no sorry 2.0.34 the newest one that was released last week those are

02:02 problematic they are going to launch a go process in the background even working around like the

02:10 import path stuff won't get around this because it'll launch this what looks like a legitimate

02:16 named thing, S-C-K-I-T.

02:18 You may misview it as scikit, but it's actually S-C-K-I-T. It's a Go

02:23 library that loads in the background. And what it does, scours your system for credentials,

02:27 posts them to a command and control server, kind of the standard play there.

02:30 And then it attempts to inject itself into other CI pipelines as a worm.

02:35 So this one's not just an exploit that's and exfiltrate your credentials,

02:40 they're actually looking to exfiltrate them, exploit them,

02:43 and then turn them into a worm to gather more and more of them, whatever projects you're

02:47 basically working on. So it traverses your home directory looking for PyPI tokens,

02:51 GitHub tokens, you name it,

02:54 and then sends them to a nefarious server someplace.

02:58 Those then locate itself and try and go again.

03:00 So if you are on these various versions of these various packages,

03:04 make sure downgrade to the safe versions,

03:08 the 2.0.33 of memory OS and 0.1.20 of the plugin and rotate

03:14 every possible credential you have that's reachable from your home directory.

03:18 Make sure you kill any SCKIT processes that are running because that is where it is talking to the command and control

03:26 server. So unfortunate news, no fun.

03:30 The write-up from SafeDep, which is linked to from the show

03:33 notes here has a really good take on basically how they got in there's a couple spots where they

03:39 don't know how they got some of the tokens they assume maybe social engineering or a leak of a

03:45 github token someplace but the various projects have put in some fixes into their ci pipelines to

03:52 hopefully prevent future attacks like this from happening and you probably could learn something

03:56 a thing or two for your own ci pipelines to harden your processes as well that is a jump scare

04:02 Yeah, less than fun.

04:04 Less than fun.

04:04 No one loves to hear that news on a Tuesday morning.

04:08 But I think everyone needs to be aware and protect their CI pipelines.

04:11 Because again, this is how they got in.

04:13 They basically used the GitHub Actions pipelines.

04:16 Not that GitHub Actions is in itself vulnerable to these things.

04:20 It just enables these kinds of things.

04:23 It's the other practices that are what made this possible.

04:26 I remember how scary it was that there were worms on the internet,

04:30 especially back when firewalls, what's a firewall?

04:33 That's what a company, that's what like giant companies do.

04:36 Yeah.

04:37 So there's, on the webpage,

04:38 they also link into the issue from the MemoryOS folks

04:43 and the OpenClub plugin repositories.

04:46 Again, I think it's a good responsible write-up

04:48 of what happened.

04:50 The repositories themselves don't have a lot of information on them,

04:53 but this article does.

04:55 Yeah.

04:55 It's just, it's scary, you know, having very cool what we can do as developers,

04:59 but at the same time, the responsibility of, oh my gosh,

05:03 something got through and it didn't just affect me.

05:06 It didn't just affect my users, but everybody who might have used something I created

05:10 now, it's like taking over all of their music.

05:12 Oh my gosh, that's a serious fire.

05:15 So. I mean, it's a common complaint. Like, why can't we just dump files into a web route like we used to and

05:19 deploy CGI applications?

05:22 Because that was very dangerous. We just didn't know it yet because of

05:25 the various vectors for exploit and attack.

05:29 Deploying software safely requires a thoughtful

05:32 and intentional process.

05:33 Yeah.

05:34 And to some degree, safety and isolation of your computer that makes

05:38 that software.

05:39 Very true.

05:39 So yeah, pay attention to a lot of the sandboxing and containers and

05:43 things like that that help you with those processes.

05:45 Would you say you got to pay attention to like tiny little issues?

05:48 Tiny things could, you know,

05:50 be an interesting way to go.

05:52 I want to talk about Tiny Mongo.

05:54 Are you familiar with this?

05:55 I am not, actually.

05:56 I'm quite curious.

05:57 My son actually was in a MongoDB hackathon over the weekend,

05:59 so maybe he might be interested in this.

06:00 This would have been absolutely awesome for that.

06:03 I mean, maybe if it was about MongoDB, no.

06:06 But I've covered this before, actually.

06:08 Gives me a chance to highlight a cool little search feature of our, if you search them,

06:12 you can actually say, only show me the episodes that exactly covered this.

06:16 So way back in 2017, nine years ago,

06:19 we covered Tiny Mongo.

06:21 So the guy behind this, Steven, I had posted way back when I had posted some issue saying,

06:28 this thing is cool.

06:29 So let me tell you what TinyMongo is, and then I'll tell you why it's back on the show after

06:33 nine years.

06:34 So TinyMongo is what SQLite is to Postgres.

06:39 TinyMongo is to MongoDB.

06:40 Okay.

06:41 In process, local file, but it goes,

06:44 I don't know, it goes a little farther than SQLite, as you'll see,

06:47 in some really interesting ways.

06:49 but it uses well-known durable backends for the most part.

06:53 So you can choose a really well-known, well-tested, well-trusted backend for this.

07:00 But basically think in process, I want something without a server.

07:04 Now, as part of this news item is very, very fast.

07:07 So Stephen looked at this GitHub issue and decided, all right, I'm going to fix it.

07:12 So the issue I filed a while ago was, this is really cool.

07:15 It supports a subset of MongoDB, but not enough.

07:19 Okay.

07:20 So I was using, when I posted the issue, I think I was using MongoEngine.

07:24 When he replied to the issue, I was using, what one was it?

07:27 I was using a Pydantic-based one, Beanie.

07:29 And then by the time he actually got the thing working, I was using just raw MongoDB queries.

07:36 But I said, look, this is cool, but I can't use it with anything based on these ORMs

07:41 because the ORMs have a certain set of startup things they do.

07:45 like ensure these collections exist, ensure this index exists.

07:49 You know, it's like testing different things as just part of the standard setup,

07:53 you know, like it scaffolds the index automatically and stuff.

07:56 And those things weren't working.

07:57 So I said, would you be willing to put in enough structure,

08:01 maybe even if there are no ops, that I could actually run a quote real application on top of tiny Mongo?

08:07 Because there's all sorts of cool reasons you might want this.

08:09 Like if you're doing a tutorial or a workshop, you don't want to have to start out and go now kids,

08:15 The first thing we're going to start is by setting up a network server and securing it.

08:18 Like, no, no, no.

08:19 Yeah, there goes the day.

08:21 You're like, okay, well, we're not doing the workshop anymore,

08:23 right?

08:24 But if you have something like SQLite, you can just say,

08:26 this is the connection string.

08:28 This is the uv pip install command.

08:30 Now let's keep going.

08:32 But it's the same program.

08:33 You just change the connection string basically to get like a real server,

08:38 right?

08:38 Like to switch over to real Mongo or whatever.

08:40 So Stephen took that idea and just totally ran with it.

08:43 and came back to me and said, hey, look, what do I need?

08:48 How about this?

08:49 You know, this last question was asked in the time of AI,

08:53 the time of really smart coding agents, right?

08:55 Just the before times and there's now.

08:57 Exactly.

08:58 So I think this actually represents a really super interesting collaboration

09:03 between me and Stephen.

09:05 So, Stephen, sorry.

09:07 So he said, well, what do I need?

09:08 I said, well, how about this?

09:09 why don't I just see if I can get Talk Python to run on top of this,

09:13 just as it is?

09:14 And I just said, all right, well, I'm going to take Mongo out.

09:18 I'm going to stop talking to the real Mongo.

09:20 And I'm going to put Tiny Mongo in and get it to run.

09:22 And there were all these issues like, yeah, it technically works,

09:25 but this thing is a thousand times slower in this way.

09:27 It's like, oh.

09:28 Interesting.

09:28 I wasn't expecting that.

09:30 Yeah.

09:30 Well, it almost worked.

09:32 Once we got it working, it was like, okay, now it works.

09:33 But it's like, these five things are insanely slow.

09:36 this type of query is not actually supported or limit doesn't actually limit it in the DB.

09:41 It pulls it all back and then it like limits it in memory.

09:43 You know, those kind of weird little, like it's fine.

09:46 But if you've got hundreds of thousands or millions of records,

09:49 you're like, no, this

09:50 is not going to fly.

09:51 This is out of control.

09:53 So we went through all those and now it's like really quite close to MongoDB performance.

09:59 Michael, would you recommend this for production usage or is it still more for exploring,

10:03 playing, teaching?

10:06 I would recommend it.

10:07 I think it's safe.

10:08 I'll tell you why.

10:09 I'll tell you why.

10:09 I think it's, I would recommend it.

10:11 If you would consider SQLite as your backend for production,

10:15 which I think actually is viable.

10:17 Yeah, it's very safe.

10:17 I think that's very safe.

10:18 It has restrictions on like parallel writers and stuff,

10:21 but I do think it's quite safe.

10:23 I agree.

10:24 If you would consider SQLite as a backend, then I think this, I would recommend this as a backend.

10:28 Yes, as we'll see.

10:30 Okay, so let's go down.

10:33 I mean, what's cool about it is the way you write is you just like write regular code

10:36 and you just import tinymongo as PyMongo, which is the standard way of creating a thing.

10:41 And you give it a connection string sort of deal that is like this file instead of this database.

10:47 And then you just write regular queries against it, right?

10:49 So what's really neat down here somewhere is all the different backends it has.

10:55 So like SQLite, it might maybe even uses the SQLite version for this,

11:00 I'm not sure, is you have an in-memory version.

11:02 So it never even writes the disk.

11:03 Like I'm doing unit tests or I'm just firing up this code,

11:08 this data, I'm going to think about it and then throw it away.

11:10 Right.

11:10 That's pretty cool.

11:11 It has this JSON backend, which is it's sort of default way.

11:15 I haven't done it.

11:16 This was not really able to totally solve the Talk Python runtime, I think, or it was like too slow or something.

11:22 But switching to SQLite.

11:23 So basically it uses SQLite with all of the SQLite transactions,

11:28 write ahead log stuff and all that kind of stuff.

11:31 And that's a really good production story.

11:33 And then look at this, Calvin, sharded SQLite.

11:35 I'm more curious about the next one, which is DuckDB.

11:38 Yeah, so you're going to ask.

11:40 Yeah, because there you've got real JSON support possibly.

11:43 Yeah, Stefan is a huge fan of DuckDB and those kinds of things.

11:47 So this is DuckDB and Parquet files as the backend, which is pretty interesting.

11:52 So if you're doing data science-y things and you were doing Mongo type of queries,

11:56 this is the way.

11:57 This is a pretty good one.

11:58 And then also it has a way to say like, you know, point that over at Postgres and MariaDB

12:04 if you'd rather have something that's kind of like Mongo,

12:06 but you don't actually run Mongo.

12:08 I mean, you literally talk to it as if it was Mongo.

12:10 Yeah, I'm unsure about that use case.

12:12 I think it might just run Mongo.

12:13 I would too.

12:14 I would totally too.

12:15 But I think the SQLite one, it's really good.

12:18 We got it really, really dialed.

12:20 So I actually linked to one of the issues.

12:22 There's a bunch of issues over there.

12:24 So I had some issues.

12:26 Steven added him and sort of referenced me.

12:27 And we just went back and forth.

12:29 And I would say, hey, Claude.

12:31 I'd open up, I had a branch for Talk Python.

12:34 I'd open it up and said, hey, Claude, check out this GitHub issue.

12:37 Can you try to implement it on top of this thing?

12:40 And he would do it and say, well, I found all these issues.

12:42 I said, okay, file some issues.

12:43 Then Steven would have Codex look at it, fix it up, reply.

12:48 And we would just pass it back and forth.

12:51 And the reason that's interesting is I'm not giving him the Talk Python code base.

12:55 And I ended up running on Talk Python training, which is like 300,000 lines of Python code

12:59 and got it working there as well.

13:01 But I'm not giving him that code.

13:02 But he was able to literally prototype both from a performance and correctness perspective

13:07 running on both those code bases by just bouncing back and forth.

13:11 Well, my AI did your spike and it found that it ran like this and this worked and this didn't.

13:17 And so, okay, I fixed this, try it again.

13:19 And we just went back and forth for like over and over for days and got it really, really

13:23 dialed.

13:24 And I thought that was, that itself is worth covering here.

13:26 Yeah, I like that workflow.

13:28 Yeah, yeah, it's very interesting.

13:30 Obviously, my prompts were like, you will not put any proprietary information,

13:35 any secrets, any source code example.

13:37 You will put all generic, you know what I mean?

13:40 But at the same time, it totally worked.

13:41 And it was really cool.

13:44 But I think this is a super interesting thing because until this recent work,

13:49 this version 1.3.1 that got released, There was no SQLite equivalent for MongoDB.

13:54 And now I think there legitimately is.

13:56 It's not 100% coverage, but it's quite high.

13:59 It runs multiple real-time apps.

14:01 Yeah, I like that.

14:02 Well, especially for the education market, being able to teach someone really quickly

14:04 without having to install and serve.

14:07 And that's a huge barrier for a lot of people.

14:09 Yeah.

14:09 It looks really good.

14:10 I'm really excited about it.

14:12 Thank you, Stefan, for doing this work.

14:13 And there might be a follow-up at some point as well.

14:16 And what a great story for open source.

14:18 Yeah, exactly.

14:19 Like a really cool way to get some actual hands-on experience on real projects by sort of ping-ponging GitHub issues back and forth.

14:28 It was crazy.

14:29 Speaking of crazy, I've heard this topic is setting the world on fire.

14:35 Well, and I wanted to bring it to light to get some people some exposure to it.

14:38 If you've not heard of Jev yet, you're probably living under a rock someplace,

14:42 and that's okay.

14:43 But the folks over at RealPython did a tutorial

14:47 of how to get started with Jev in Python.

14:50 For those of you who don't know, Jev is a model from a company called TypeSafe AI

14:55 that answers typed results like yes, no's, probabilities,

14:59 scores, and pick from options instead of you talking with it.

15:03 So you don't chat with Jev.

15:05 You don't come into Jev and say, hey, build me an app or tell me a bear story

15:09 or whatever the thing you may want.

15:11 propose to JEV sets of information, evaluation,

15:16 the kind of result you want. And then JEV can, in parallel,

15:19 evaluate those things, either scoring them, giving them yes-nos.

15:23 I mean, the docs from the tutorial are basically a customer service example where the customer says,

15:29 I've tried contacting support like three times. Can I please get escalated?

15:33 And so then the JEV for that basically is, here's how you know if the person is telling the truth or not. So they have a

15:39 YesNo or TrueFall.

15:40 And there's some custom various data types that go along with it.

15:43 So what's nice is that the real Python article will get you started with Jev.

15:48 So you'll understand the data types.

15:49 You'll understand like what Jev's about, that you can't just chat with it.

15:52 But I think, and the buzz is real.

15:54 Like I've heard tons and tons of people talking about it.

15:57 It looks like the invites are back open again.

15:58 I was actually able to sign up for Jev this morning

16:00 and try it out, but it is proprietary.

16:04 Jev is a hosted closed wait model.

16:07 There are no waits to download.

16:09 There's no self-hosting of this thing.

16:11 Everything called OpenJev is an independent implementation and not TypeSafe's model,

16:18 which means your data leaves your machine.

16:21 So I don't know how much you trust TypeSafe AI.

16:24 I don't know who they are, where they're hosted, what their business models are.

16:28 Obviously, it's not free.

16:29 You have to put a credit card down to use this model and try it out.

16:33 But every call sends your input to the input text you put in there to a third party.

16:38 So be careful, like much like you just mentioned

16:40 how you are sanitizing your inputs for working back and forth on the tiny Mongo project

16:46 with Stefan, be careful what you send into this one

16:48 because you don't control it just as much as you control

16:51 what you send to Anthropic or OpenAI.

16:54 Those are pretty well established companies with a good business model and they seem to be running on,

16:59 you know, credibility and reputation.

17:01 You have to do your research and know what you're gonna be sending over there.

17:05 So think twice before sending customer messages,

17:07 tickets or anything sensitive.

17:08 The other thing is like costs and stability are a question,

17:12 right?

17:12 So basically Jev bills themselves as very cheap and it is very cheap. Like you can,

17:17 and very fast, you can have it sort through.

17:20 I heard someone tell me an example, like look through my inbox of a

17:23 million messages, categorizing them for like needs immediate attention,

17:28 is about a customer, et cetera,

17:30 et cetera.

17:30 And it can build you a table and sort that table in just milliseconds.

17:34 It's that fast.

17:35 I want to give people an option here.

17:37 It's a good practical intro.

17:39 It shows how to use the null, the score, the choice primitives,

17:43 but there are open options out there.

17:44 I know I keep caveating this like episode,

17:47 this, this section of the episode.

17:49 There are some open options to look at instead.

17:52 For example, Jevk5, it's an Apache 2.0 licensed weights and code.

17:58 So that's kind of my front runner right now. I've not tried it,

18:00 but it does accept type safe style requests.

18:04 These are API compatible, but they are not the JEV model. So you're going to get different potential results.

18:09 There's also one called SimF, which was formerly called OpenJEV. I'm sure lots of takedown notices

18:15 came flying back and forth as all these alternatives came out that were also called

18:18 something something JEV.

18:20 Another one's OpenJEVSGLang.

18:22 Again, the catch here is these are not the JEV model. They are close.

18:26 There was actually a really interesting JEV in 25 lines of Python,

18:31 on a little bit tongue in cheek post that also came out about this.

18:35 But what they did here was use the Quinn three model

18:38 and told it to act like Jev.

18:40 And for most things, I think it gave reasonable results,

18:43 but also at the very end, this was a real tongue in cheek post

18:48 kind of poking fun at the hype that is Jev right now.

18:53 So I wanted to put that out there for folks to be just mostly aware of the fact

18:56 that you're sending your data to a proprietary company

18:59 and be careful what you're sending to it totally trust it and have a business agreement with that organization beyond putting your credit

19:06 card into a credit card field.

19:07 So I hope folks are vigilant as they go forth and try these things.

19:11 The hype is real.

19:13 Oh my gosh, you couldn't again move last week without hearing about Jev in some

19:17 meeting and some article and some topic.

19:22 It's a very interesting way and a very interesting usage.

19:24 It's a very, I think it's a great pattern for reducing token spend and usage on other models

19:31 collaboration or combination with other models,

19:33 but you need to understand what it really means

19:35 in action.

19:36 It's a super cool thing.

19:38 You're right. It's absolutely blowing.

19:39 Jev is blowing up.

19:41 I had to watch some YouTube videos this week.

19:44 I'm missing something because there's a lot of this thing I've never heard of going on around and around.

19:49 When you sign up for a Jev account, it makes you take a pop quiz to ask you whether or not you can chat with Jev. I won't give you the

19:56 answer because if you don't know, you shouldn't probably be using Jev.

19:59 If you got to ask, It's not for you.

20:01 Yeah.

20:01 I was going to say that there's never been a time, I think, where we're sending as much detailed information to other companies and other places as the last couple years.

20:14 Be careful out there, folks.

20:16 It's real.

20:18 Yeah.

20:18 But the reason we're doing it is because it's so productive and so useful, right?

20:21 It is.

20:22 I mean, the patterns, I think,

20:24 in looking at some of those open wide options, if you can put in place, again,

20:28 I'll stress, probably why I stressed things last week.

20:30 was around evals.

20:31 Building good evals for your CI pipeline to run against these models,

20:35 whether they're classifiers like JEV or whether they're LLMs like the QIN or other open weights models,

20:41 means you can swap back and forth with confidence.

20:43 And if you can do that, you can use one of these

20:45 open weights models, open source,

20:47 even versions of these models with more confidence on your own

20:50 GPUs. These things will run on even very small CPU,

20:53 GPU.

20:54 I think there was a QIN 306B model that they're using in the parody article that runs on like a raspberry pi you can get those anywhere

21:02 nice but i'm not using Anthropic for something i've been using glm53 flash and that's been

21:08 super super neat but i want to talk about an age of innocence that may be over here this is an

21:12 article i don't typically cover articles i typically more but but it kind of pulls out it highlights a

21:17 really important thing that the reason i pull it up is because it kind of broke my understanding of

21:22 Python performance tips for one particular axis.

21:26 So this is by Timothy Ivankov.

21:30 So cool to write this up.

21:32 I think it might have had a little writing help, but that's okay.

21:35 So here's the headline.

21:36 It's reading Dunderdict once, one time.

21:40 Reading Dunderdict of an object permanently de-optimizes attribute access.

21:46 What?

21:47 Why is that?

21:48 Why is that not good?

21:49 And I was a wonder's point of this comes from PyCoder.

21:51 The permanence of it is the striking bit.

21:54 Permanently, for the rest of that object's life cycle,

21:57 its performance is broken.

21:59 Okay, so here's an old performance tip.

22:02 And I'll tell you the one that I used that I thought was amazing.

22:05 And it was amazing.

22:06 If I have a loop and in this hot, let's call it a hot loop here.

22:09 This is an example.

22:10 A million times around, we're just going to do the same thing.

22:12 Like we're going to say, create an attribute,

22:14 self.value, access self.value in the loop,

22:17 increment plus equals on the self.value.

22:20 Well, traditionally, that would go into the DunderDict, find the value,

22:24 pull that value out, change it, store it back into the DunderDict, right?

22:28 That was the backing store for its fields, which is weird,

22:32 but that's how it worked.

22:33 So here's what you do.

22:34 Instead of doing that interchange over and over and over,

22:37 do it on a local variable.

22:38 Create a value, do all your work on a local variable in the function,

22:43 and then at the end, set the value to the class,

22:46 right?

22:47 Create a value, and then eventually say self.value equals value,

22:49 right?

22:49 Super simple.

22:50 That used to make a big, big difference.

22:53 And now it does still a little bit, does a little bit. So it technically works.

23:00 However, if you do this enough times,

23:03 it turns out that the new specializing adaptive interpreter notices that and it drops the load adder,

23:10 which is the bytecode instruction that reads the field from the dictionary.

23:15 Eventually it drops it and it starts processing it different,

23:19 right?

23:19 And so what it does is it actually starts to look at just,

23:23 well, where in the offset, like a fixed byte offset

23:26 into the object's storage for this, which is like,

23:29 wow, okay, pretty cool.

23:31 So that's, I think, since 13, really interesting.

23:34 And you can see the different aspects.

23:36 It turns out that it's even more significant, this change of the problem they're suggesting or pointing out.

23:43 Is that really a problem?

23:44 It's just breaks in optimization.

23:46 that the breakage is stronger in a free threaded world.

23:49 And the article goes into why.

23:50 So check this out.

23:51 If you just say, like maybe this is like a debugging thing or whatever,

23:54 you just say print dunder dict of the object and then you go have that function run.

23:59 All of a sudden it's 50 milliseconds instead of 30 milliseconds.

24:03 1.5 times slower.

24:04 What do you think?

24:06 It's intense to get down to that level.

24:09 It is, but there's a bunch of little simple things that you're like,

24:13 I could read the fields or I could just say star star dunderdict and do this and that.

24:19 And it turns out that that actually, it didn't used to make any difference,

24:23 but because of the specializing adaptive interpreter,

24:26 now it does, right?

24:28 So like if you just ask vars of an object or if you ask if a field is in the dictionary,

24:34 well, here's the really tricky one, a shallow copy,

24:37 right?

24:37 Like, well, that's a totally reasonable thing.

24:38 I'm going to clone this before I hand it back or something.

24:42 Well, that clone that you did, it can attribute access.

24:45 to access it yeah it makes it 50 slower i've definitely done that mostly when i'm debugging

24:49 or triaging code or kind of walking through the interpreter because it's easy to access i can't

24:55 think of too many times where i would have tried to do it in code you could see how it's a it's an

24:59 easy workaround it's like oh it's just right there i'm going to reach for it yeah exactly so these

25:04 the the first three like are kind of um debugging ones but this last shallow copy this is a

25:08 legitimate thing, right? That you might do.

25:11 And so here's the, this is the part that I didn't get because the world has changed and I haven't been like prototyping everything at that level.

25:20 I'm a big fan of dunder slots.

25:22 It means you can't dynamically add stuff to your class,

25:24 but more importantly, used to mean that this whole dictionary mechanism was no longer used.

25:30 And it basically used an offset into a list that was just in each instance.

25:35 So it honestly made things a lot better in terms of memory and in terms of attribute access was significantly faster.

25:42 Well, apparently slots makes no difference anymore because the specializing adaptive

25:48 interpreter, at least in this use case, right? This is like a,

25:50 I'm accessing the same thing a lot of times, right? It could be, you know, profiling is tricky.

25:55 Performance is tricky, but basically the specializing adaptive interpreter for hot pass slots versus not didn't really matter.

26:04 And I didn't know about slots until I read Luciano's book,

26:07 Fluent Python.

26:08 That's where I discovered that.

26:10 And now you're saying it really doesn't matter anymore because of the new bits in there.

26:14 Yeah, exactly.

26:15 So I believe I might have learned about it as well from Luciano's book.

26:20 So yeah, 3.11 and beyond, there's a lot of the faster CPython thing.

26:25 Changed some of these things.

26:26 And it changed how this, especially with the specialized and adaptive interpreter,

26:30 changed a lot about how this dictionary and access, and that's the whole story,

26:34 slots, dictionary access, optimizing or not optimizing it, and so on. So I thought this was an interesting

26:41 dive into something that's, you know, everyone uses classes,

26:45 objects, even if they're not big OOP folks,

26:48 you still create objects.

26:50 I mean, do you have a number?

26:52 You know what I mean?

26:53 Well, it's just, and it's convenient,

26:55 but I think if you were, again,

26:57 performance debugging and reaging and you ran into this,

27:00 you've been surprised. Yes, I would. I would certainly say

27:03 so. I would say so.

27:04 Michael, would you want an interesting way to find out what's new or what's

27:08 newt in Python?

27:10 So I got this extra here I wanted to share.

27:13 This is cool.

27:14 I've seen this now. I thought I had not seen. I've seen this is cool. This is a dungeon crawler called what's newt.

27:21 And it, but what it does is it runs you through the what's new in Python.

27:24 And this version is specific for 3.15.

27:27 You can't run it with prior versions because it actually uses the features in

27:31 3.15 to check that you've completed the challenge in the dungeon.

27:36 So I've got the dungeon up right here. So I'll just read off real quick. So this one starts off with the first room of the

27:41 interpreter and the one nobody sees.

27:44 Slips of paper are wedged into every crack of the masonry.pth

27:48 files.

27:48 Hundreds of them, each one adding a directory to the path.

27:52 A few begin with the words import.

27:54 And those have scorch marks, rightfully so.

27:57 A brass plaque is bolted to the wall. A new file

28:00 sits on the lectern beneath it, unwritten.

28:03 And so at this point, you have a puzzle,

28:04 which is what runs before your program does.

28:07 And it has a link actually in the terminal.

28:09 That is a hyperlink.

28:11 It's a nice little TUI app.

28:13 And when you click on that, it'll open up the related PEP to basically

28:17 give you a hint for it. So if you want to solve it, you just type solve.

28:20 I'll type solve.

28:23 It'll my butt because it does not have VI key bindings in there. So I had to like use my arrow keys.

28:28 You probably had almost just like a caveman.

28:33 And so based on the PEP 829, it added in the new

28:38 namespace, a callable syntax for bringing in a callable into your PTH file or the new

28:47 sys.path, for example. So they're deprecating the word import.

28:51 And now 315 adds this file format that can only name a callable.

28:55 Here's how you write it. And when you click check,

28:57 it's like solved. A reader can now see what starts up without executing it. The site modules batches every static

29:03 sys.path extension first, and then only runs the startup callable so that a.start file can rely on

29:09 the path being complete.

29:10 Please press enter when you're finished reading.

29:12 And then it takes you into Do you want to go north,

29:15 south, east, west on your journey and learn more?

29:18 And tracks your progress so you can see what scores I've gotten,

29:21 how many puzzles I've solved.

29:22 If you go east, it's a very lazy path.

29:25 Yeah.

29:26 So that's super fun.

29:28 I've not spent enough time reading the what's new in newer versions of Python,

29:33 kind of like you alluded to in the last article,

29:35 because we just use Python.

29:37 And a lot of times I'm not leveraging some of these new features, but this is a fun way

29:40 to learn what they are.

29:41 I figured I'd throw that in there.

29:42 Oh, it is very fun.

29:43 This is from Barry Warsaw, right?

29:45 I believe so.

29:46 That's a good question.

29:47 I believe so.

29:48 What's new?

29:49 It is.

29:49 You're right.

29:50 It is Barry Warsaw.

29:51 Yep.

29:51 Good job, Barry.

29:52 Nice work, Barry.

29:53 I love it.

29:53 I'm here for it.

29:54 And especially with the hype around Dungeon Crawler Carl right now, this is very appropriate.

30:00 Brings me back to the days of MUDs.

30:02 I used to play multi-user dungeons.

30:04 I mean, me and my friends, we would spend tons of time in these places back in high school.

30:09 People got to remember when I was in high school, there literally was no web.

30:14 Not like it was bad.

30:16 We're dinosaurs roaming.

30:17 The worldwide web, I know.

30:19 The whole thing was created in 1993 when I was in college.

30:23 So before that, you're like, well, we got Telnet.

30:26 What are we going to do, folks?

30:27 Yep.

30:28 Muds.

30:28 That's what we're going to do.

30:29 We're going to use Gopher.

30:30 It was a lot like this.

30:31 Yeah.

30:31 Gopher, Archie, Telnet.

30:33 Oh my gosh.

30:34 News groups.

30:35 Yeah.

30:36 Get your newsreader out.

30:37 What's so ironic is it was so amazing.

30:40 And it seems so futuristic.

30:42 I bought things off of news groups, like going in through Telnet and then getting the news

30:47 groups.

30:47 And I bought an expansion card for my HP 48 GX.

30:51 Incredible.

30:51 All right.

30:52 Well, from the past to the future,

30:54 well, to the present, which feels actually legitimately

30:57 future.

30:57 Tell me about the future.

30:58 Things used to be so simple.

30:59 You mentioned throwing just a set of files to a folder and you may be behind a C panel

31:05 so you can admin your website.

31:07 Oh, that was almost too modern for me, C-panel.

31:09 Yeah.

31:10 Yeah, just FTP it up.

31:11 I'm sorry.

31:12 Not even SFTP, just straight FTP.

31:14 Just open FTP.

31:17 So I want to come back and do one more Kyle Lintit because I couldn't help myself,

31:23 Calvin.

31:24 I saw, oh my God, I saw this one when we talked about the big data video.

31:30 So here's more homework for folks.

31:33 This is the perfect encapsulation of just how stuff gets so complicated for such a simple thing.

31:39 So here's this 10-minute video.

31:42 It tells the story of multiple mini personas throughout

31:46 that we experience as software developers and data scientists and so on,

31:49 trying to build tools, just a simple thing.

31:53 You know, it starts with, okay, I need you to build a dashboard

31:57 with a single button that does, I don't know,

32:00 you know, whatever the button does, right?

32:02 Pretty soon, there's this guy over the shoulder and says,

32:06 what about what are you doing with a regular database you need a graph database he goes i

32:11 don't need to have a graph database for 200 000 users he goes what if we had 200 000 200 000 and

32:18 one users he goes gosh his delivery is genius in this one it's so good so good my highlight was the

32:26 pearl guy pearl pearl you know devil shows up on his shoulder yeah it does yeah there's this this

32:33 other guy's like well we need a ddd you know domain driven design bounded context for the

32:38 button we need docker builds for the button it just goes on and on and just oh it is absolutely

32:46 good and the pearl guy somewhere farther it is like i could have built this with a single pearl

32:51 script and a cron job back in you know whenever but you portray them perfectly i mean if you've

32:57 been around any amount of time in this world you're gonna you're gonna laugh yeah yeah yeah it's it's

33:02 So good.

33:02 So everyone, I leave you with a delightful,

33:08 too realistic look at how software is built these days.

33:10 Yeah.

33:11 Well, this was really, really a good episode, Calvin.

33:14 Like you said.

33:15 Yeah.

33:16 Catch you later.