Lightning Talks (1/2)

Speaker 1 [00:08]

Wonderful. Welcome, everyone. Welcome to one of the most unique session of this conference, not to brag, Lighting Talk sessions. So for those of you not familiar with Lighting Talks, Lighting Talks are for bold ideas, bite-sized brilliance, and community energy. Five minutes of your time where you can speak about anything Python-related or not. No Q&A, don't expect question or answer, just the good stuff. When and where this conference will happen? Well, this session, duh. And tomorrow we have another session, I think slightly longer, one hour and a half, from 15 to 6, up until 2, or 7, a quarter past 7. All right. Okay, this is for the boring part. Let's talk about what's on today's menu first we will start with some event announcements those talks will be limited to one minute and for everyone submitting an event announcement today please come to the front rows so that you will come on stage one after the other I don't think we need to to name you just like show up and do your talk and then we have some playful chaos I will tell you about that after the event announcements. I just want to highlight I've been called out on publicly on the internet, not to be trusted. One of the reason why I'm sharing this slide is because there's also Tania featured in that feature in that picture since we're live. Tania, we love you, we miss you. She was very kindly supporting a very similar session we did together last year on Lighting Talks. This year though, I have the wonderful Johannes joining me. A huge A huge round of applause to Johannes, please. With Johannes' help, we do the lightning talk selection, and also I'm going to tell you afterwards what are going to be the playful chaos we'll play together. But just FYI, we're going to use this timer today. We'll try to use it. If it's too noisy, we're not going to use it. But at the end, this is the timer for the end of the talk. All right, we have the gong. Just to give a preview of the playful chaos I'm talking about, I just want to talk about the bingo game we're going to play together. And this is a great opportunity for everyone in the audience who wants to play the game to grab their bingo card. If there's anybody here not having their card and still want to play, just stand up, please, and somebody will bring one to you along with a pen. All right? For all speakers, all potential speakers happening today or tomorrow, please mind that we do have a list of entries you have to use you must pick at least one you can pick as many as you like and feature them during the talk during the your presentation some ideas ask can you see this or like can't read the code or complaining about errors cracks a dead joke or forget to plug in forget to share then we have some like Harry Potter inspired jokes about windows or different operating systems reference life coding says hoops says one last thing and uses a walrus operator that's a very good one so any of these this is not complete list 150 entries there you can pick as many as you like and feature them during the talk and people attending live should pick the boxes and and then I'll tell you what happens if you win all right let's do event announcements now and since I am the host of this session I I will start, and I would like to start by telling you, do you know where Python meets pasta, passion, and powerful code? Any guess? Well done, it's actually in Bologna, Italy, at the end of May, next month, for PyCon Italy 2025, which I'm proudly introducing you. Bologna is the city of portico, strotolini, and tech. Come for the code, stay for the comms. Bologna 101 is home of the oldest university in the world, Capital of Italian cuisine, that's very true, I can vouch for that. City of Towers, actually the two towers, the Lord of the Rings fans here could be amused, and handmade pasta, and we do have fast Wi-Fi and very slow food as well. 100 and plus inspiring talks, more than 10 hands-on workshop, three social events, friendly community, and yes, real espresso, not coffee. So, join us, CodeConnect, Carboload, Pykin Italy 2025, grab your tickets, pykin.it. And that's all I had. Please come the next event announcement, come on stage, who wants to be next? We need to go very quickly, so it's one minute at a time. All right, let's do it. Well done. I like the attitude. I can take the time and you do the gong thing. Good. Mm-hmm. Is it popping up? It's happening. Yes. Okay. Okay. Go ahead. Is it popping up? It's happening. Yes. Okay, go ahead. All right. Okay. Can we have it displayed? Yes. Yes. Great. Hi, everyone. I'm talking to present Pydata Berlin. We'll be on the 1st to 3rd of September.

Speaker 2 [05:55]

So yeah, we have a

Speaker 1 [05:57]

Call for proposal that will close on the 4th of May and yes

Speaker 2 [06:00]

And yeah, if you want to come.

Speaker 1 [06:01]

if you want to come for Berlin.

Speaker 2 [06:02]

for Berlin, for the...

Speaker 1 [06:03]

the conference and the clubs.

Speaker 2 [06:05]

and some techno

Speaker 3 [06:09]

we are having a

Speaker 2 [06:10]

having a call for paper

Speaker 1 [06:10]

and call for papers open.

Speaker 3 [06:11]

the papers open.

Speaker 1 [06:12]

that

Speaker 2 [06:13]

We want to fill our form.

Speaker 1 [06:14]

our 40 talk slots and bear with me I think nine tutorial slots and we already have a couple of keynote speakers. That's it. I really hope we can use this gong once. Yep. All right. Take it away. Doesn't react.

Speaker 2 [07:00]

Why doesn't it work?

Speaker 1 [07:04]

Second to connect I don't it technical difficulties. Okay. Okay, sir. I mean I'm too impatient. All right Something weird is on the screen And you have 15 seconds 12 10 8 5 For there's no there's no external one. That was your sci-fi. Ladies and gentlemen. No, it's That's about noise It's no I'm just play shall we crack that jokes Yeah, I need to look for it Okay, I don't want responsibility. Let's ask Judge EBT to crack a dead joke. Okay. How did the barber win the race? He knew a shortcut. That's terrible. What happens when a frog's car breaks down? He gets towed. Can I please stop Telling dad jokes Two guys walk into a bar The third one wasn't The third one Yeah That's nice actually Can we probably try tomorrow Okay so Then I have to Imagine there's a nice slide here About your sci-fi So your sci-fi is a European conference for scientists that work with Python. It will be in August 18th to have to check myself. The 18th of August we'll start this ZeroSciPy and we'll be in Krakow, Poland, so not so far away from here. The call for papers is still open until April 28th so the Monday you can still get a chance to get your talk in. We do have two days of talks, two days of tutorials, and we have one day of sprints. And there will be a lot of maintainers of a lot of

Speaker 2 [09:23]

of a lot of interest.

Speaker 1 [09:24]

interesting Python packages, NumPy, We'll be right back. Pandas, other people that are maintainers of well-used packages, so you can meet them. And also, if you would like to become a sponsor, there are still some sponsor packages available. And the URL would be euroscipy.org. Thank you. All right. I love you. I love you really. Okay. Yeah, fantastic, now we have paid it to Paris. Absolutely, so I was thinking of making a slide, but I'll go through it fast, so.

Speaker 2 [09:58]

so you don't

Speaker 1 [10:00]

To watch any slide or anything Paris you have to come to Paris. It's 30th of September 1st of October

Speaker 2 [10:09]

as many slides as Valerio to tell you that we...

Speaker 1 [10:10]

you that we eat well in Paris as well.

Speaker 4 [10:14]

So come join us.

Speaker 1 [10:16]

You can submit until Sunday if you haven't done it. Why? Yeah, and that's the end

Speaker 2 [10:22]

Yeah, and that's the end of the extension, so that's the...

Speaker 1 [10:24]

so that's our deadline. Please join us. That's all I have to say. Lovely. Well done. This is a great reminder that you don't need slides to do lighting talks, by the way. Yeah, especially for

Speaker 2 [10:38]

Especially for someone who has.

Speaker 1 [10:38]

someone who has like failed two times to connect her computer today so I'm not going to try for one minutes lighting talk so what

Speaker 2 [10:48]

Why? I want to tell you.

Speaker 1 [10:49]

you about something is because who know about Humble Data? Not many people know that we have Humble Data yesterday. Only a few people. Yeah, so unfortunately, our event, I can't say it's a successful event yesterday. I'm so sorry that it doesn't work out. You know, the organizer provides such a nice venue, all the food, but we have only one person show up yesterday. So.

Speaker 2 [11:22]

So what is Humble Data? Humble Data is kind of like Django Girl, so it's for absolute...

Speaker 1 [11:26]

So it's for absolute beginner, but instead of teaching them how to build a website,

Speaker 2 [11:28]

how to build a website.

Speaker 1 [11:29]

learn about data science. They can learn about pandas and, you know, psychic learning.

Speaker 2 [11:33]

MyKitLearn and MapoLib, all that stuff.

Speaker 1 [11:38]

that next year someone can organize it. Thank you. Thank you. That's lovely. All right. Are you doing the timing? That's brilliant. Yeah, that's enough. Thank you. Okay, let's hope for the best. Yay. All right. Okay.

Speaker 2 [12:11]

Johannes and me would like to invite you to the Python Bar Camp in Karlsruhe.

Speaker 1 [12:15]

It's an event.

Speaker 5 [12:16]

It's an event.

Speaker 1 [12:16]

much smaller than here, it's more based

Speaker 2 [12:19]

It's more based on knowledge sharing and experience.

Speaker 1 [12:21]

And experience exchange and the topics is like in a bar camp.

Speaker 2 [12:24]

developed in the session as well or in the morning.

Speaker 1 [12:27]

or in the mornings. So please just join us.

Speaker 2 [12:30]

It will take place in February.

Speaker 1 [12:32]

in February or March.

Speaker 2 [12:34]

So, who is interested, just contact to us or...

Speaker 1 [12:40]

Look at the home page. There will be

Speaker 2 [12:42]

There will be an update.

Speaker 1 [12:42]

an update.

Speaker 2 [12:43]

if we know the dates.

Speaker 1 [12:47]

Thank you Hi, I'm the odd one out because to be honest

Speaker 2 [13:16]

be honest this doesn't have to do anything with python but it's a bar camp like the previous one

Speaker 1 [13:20]

We are organizing a freelance.

Speaker 2 [13:22]

organizing a freelancer bar camp in Hamburg,

Speaker 1 [13:23]

In Hamburg, the first one

Speaker 2 [13:24]

first one ever

Speaker 4 [13:25]

possible topics are

Speaker 1 [13:26]

Topics are, as always, at the Barclays.

Speaker 2 [13:27]

as always at the Barcamp, what you bring.

Speaker 1 [13:29]

you bring. It's on the

Speaker 2 [13:30]

on the 26th of july and it's in the speicherstadt in hamburg it's close to the

Speaker 1 [13:35]

close to the Elbphilharmonie, we'll have nice talks, nice food.

Speaker 2 [13:39]

and we would welcome any questions.

Speaker 1 [13:40]

Welcome anyone

Speaker 2 [13:41]

if anyone wants to be there.

Speaker 1 [13:41]

Thank you. Well, then, folks. Let's see your windows machine.

Speaker 2 [14:07]

It works without problems. That's unexpected. Yeah. But, go ahead. So, this is the meta announcement talk. Basically, I maintain a website, which is the largest archive of Python conferences and Python conference deadlines. Why is it the largest? I stole all the other ones. You can see cute snakes in the bottom. You can see all the deadlines to FlaskCon. We heard about PI data Paris right here and you have a calendar overview here You have a map right here, and we can see we need much more in Africa as always If you have event series if I have at least two of you these automatically go here So PI con Germany if you want to scroll all the way to the back and even see like the archived Thingy of this this is going to take forever So we're not looking at it, and if you have if you're organizing stuff I actually this is all open source the data is open source, and here is a nice Give how you how you add your cool event? Yes, I've been dead lat Lin dot es because I'm funny like that

Speaker 1 [15:31]

All right, yeah. I like that. All right. So, I just don't want to get too much of your time and give speakers the chance to come on stage as soon as we have that connecting. Brilliant. All right. Let's talk about the playful chaos I wanted to introduce. Right. All right. So lighting and I have Johannes helping me out with the showcase. Lighting talk sessions, we have, as I said, in the opening this morning, we have prizes from the speakers and people in the audience. And we have speakers there that I'm going to present and also surprise challenges for speakers. So, first prize for the best lighting talk, I'll tell you later how do we announce the best talk, is this amazing Python fluffy, pushy thing. This is really cool. Now I look like a wrestler from the 90s, so let's not do that, thank you. And the first prize for the bingo, in case anyone does it, across the two days by the way, is this amazing hat. It's actually programmable, it's really cool. So I would recommend that. All right. And if we have more winners, we also have more things to give away. So don't worry about it. Let's talk about the speaker dares. So the deal with the speaker dares for today is that we have that box. Any speaker is free to choose how many dares they want. So pick one, pick none, pick all of them, up to them. So first one is the bunny hat. This one is actually very cool. You can use it as a pointer during the talk. So you can point lights with that and you have to wear it of course during the talk. No further explanation. Another one is the talking puppet. So the deal with that is that you have to hand it in one hand and use the other one speaking to your talk. So that is the dare. That's very cool. If you're already wanting that, unfortunately it's already booked, so it's for Veronica, so sadly you won't have it. The third one is the fantastic eyes of heart. So if you're feeling lovely and passionate about your project, you should definitely wear those, and having to do the talk. The other one is the Queed hat, which is also very nice, and beautifully wore by Johannes with Proud, yes, you should probably do this like that, yes, that's nice, and last but no means least, we also have, well, there's also another one, we have bubbles, so if you fancy doing bubbles during your talk, this is what you can do, right, randomly during the talk, do bubbles, last but not least this is a game we played many times and it was actually lots of fun this requires audience participation if the speaker is going to pick the rubber duck what we're going to do is we select a couple of words from their talks all together and as soon as the speaker says one of those words we're going to quack together right imagine the the the word we chose is speaker right and I say speaker and you do exactly even louder exactly that's the attitude brilliant all right and finally this is the last thing I want to say and then we we pick up the first speaker well before introducing the random challenge card let's speak let's have the first pick up so they can get ready in case all right so we're gonna announce the first speaker and then but we actually pick two, so the first two can get ready. The first one coming on stage will be, oh, the rule is, if you're not in the room, we're going to select the other one. Or if you don't want to call me for any reason, that's completely fine. Right, go ahead.

Speaker 2 [19:35]

The first one is John Robert with Do Not Fear AI.

Speaker 1 [19:39]

Do we have John Roberts in the room? John, are you here? Brilliant! Fantastic, John! Okay, so John Think of the dares you want to play in case you want, of course, no pressure And the second one to be on stage will be

Speaker 2 [20:04]

Franz Kiraly, German Center for Open Source AI, the new and biggest German OS umbrella nonprofit.

Speaker 1 [20:10]

umbrella non-profit. All right, Frank, are you here? I am here. Oh, brilliant, okay, you're already there. Okay, fantastic. Okay, so while they're both thinking of their dares, I just want to explain the random challenge cards.

Speaker 2 [20:23]

cards.

Speaker 1 [20:24]

This is up to the speakers what game with this no pressure the thing is that Speakers can choose how many cars they want depending on how bold they feel During the talk it could be zero could be as many as like and it's gonna be like random challenge Really randomly selected and the cool thing about this this thing is that we have so many That I would like you to ask if you like the idea if you like playing the game together you can come tomorrow at the booth and propose your own cards so I've prepared 30 around 30 around here and I give you an example for example one random challenge could be pretend to be incredibly afraid of one of your slides it's more like an acting field or like begin your talk or your next slide like you're narrating a dramatic movie trailer this is the kind of challenge we're facing today so this is up to speakers to decide whether they want to do the challenge and how many they have all right so that's the game we're going to play today tomorrow we're going to play different games probably and let's have fun together so John come on stage? Use a round of applause, please. All right. You have five minutes. Okay. Pick your dares, if you like, and do you feel bold enough to try a challenge card?

Speaker 6 [21:49]

Yeah, I'm trying.

Speaker 1 [21:55]

Fantastic choice. Well done. Pick one card. So you choose your destiny. I feel, I feel, I feel like gentle today. If you want to change that, you can do that. Just because you're first.

Speaker 2 [22:15]

Let me change that. All right, you change it.

Speaker 1 [22:22]

That's your last choice, okay, all right

Speaker 6 [22:26]

Whisper your first sentence, then shout the next one.

Speaker 1 [22:30]

Okay, did you get that?

Speaker 6 [22:33]

Whisper your first sentence, then shout the next one.

Speaker 1 [22:36]

one all right whisper the first sentence then shout the next one all right it

Speaker 6 [22:37]

Alright.

Speaker 1 [22:42]

could be and it could be any time in time

Speaker 6 [22:56]

What is it?

Speaker 1 [23:01]

They say Mac plug-and-play

Speaker 6 [23:04]

Yeah, but I don't know what it's like.

Speaker 1 [23:07]

No, I think it's it's setting things. Let's try again. It's a display. Yeah, try allow Now this should be working It's a bit fussy, you know Yeah, there is my show Is very good. All right, so you're right Okay, any time actually so temperature

Speaker 6 [23:31]

So, just to start, what technology do you fear? I'm asking a question. What technology do you fear? What? Pizza. Any other person, what technology do you fear? Okay, I think I'm going to answer the question. A lot of us fear AI. When you ask everybody on the streets what do they fear, they fear AI because they think AI is going to take their jobs, AI is going to destroy the world, AI is going to replace humans, AI is going to do what...

Speaker 1 [24:10]

Whatever

Speaker 6 [24:11]

And, that's why, so this is an example, we see AI as a competition, but instead, AI is supposed to be an assistant. So, and that's where this comes in. So, I started this website called DontFearAI.com, and now it's still like a newsletter, but I want to extend it. and what Don'tFearAI.com is supposed to do, it's going to... So I'm someone that explores a lot of technology, so MCP, agent-to-agent, AI agent, and things like that. Last year, I spoke in this conference, I spoke about the security of AI, and I do a lot of things with AI. Do I work? Presently, I don't do so much with it, but in my free time, I do a lot of things with it. and what I want to do is just to explain basic concepts like things you see recently and things you... because I've been in the industry since like 2018, worked a bit with self-driving cars and a lot of things in AI and I see that a lot of people have misconception about AI and that's why I have this newsletter hopefully I may start a YouTube channel if I have the time but just going to explain and also build a lot of projects. Some of the things I've done in the past last maybe two three months so I viewed an open source of this open deep deep research so deep research is from open AI and then you have also from entropy and everyone so I did an open source open source version of it did a couple of project with AI agents and spoken in a lot of and participated in a lot of hackathons so yeah so that's what I want to show you my name is John Roberts I work at Sonics it's one of the silver sponsor of this conference yeah that's what I have to say so for the question I say whisper the first sentence I was it's gonna be okay when we start when was pi data founded when once pi con DE founded? 2011. Are they right? Yeah, 2011. Yeah, thanks so much.

Speaker 1 [26:53]

All right, we've put back the cards, so Frank already has chosen the drops. Frank's. Oh, Frank's a toy. Apologies. We're going to pick the next speaker. We're going to pick the next speaker. While Frank's get prepared.

Speaker 3 [27:06]

That's a bingo thing, by the way.

Speaker 1 [27:08]

It is All right Interesting hmm, yeah props yourself. Yes

Speaker 3 [27:17]

Yes, yes, yes, yes.

Speaker 1 [27:18]

Okay, you know you have five minutes, right?

Speaker 3 [27:19]

I know you have five minutes, right? Situational comedy, yes. Yeah, go ahead. I'm trying to find the...

Speaker 1 [27:24]

Do we have Nico in the room, UV and PEP 751? Do we have it? Nico, are you here? Brilliant, okay, Nico's coming. Get ready. So where do I connect? Do you feel bold of trying some random challenge card? Yeah, sure.

Speaker 3 [27:43]

How many do you have do I have to do all of them or what's what's the punishment or not?

Speaker 1 [27:50]

I mean, as many as you chose.

Speaker 3 [27:51]

So where's your HDMI cable here?

Speaker 1 [27:51]

So where's your... that you can just HMI.

Speaker 3 [27:57]

Yeah, I need HDMI Can you see this? Can you see this?

Speaker 1 [28:06]

Sorry for destroying your...

Speaker 3 [28:08]

Exploring your equipment. Is this already my time running? Yeah, yeah. Oh, no, no, no, no. Oh dear.

Speaker 1 [28:15]

So, choose your challenge.

Speaker 3 [28:17]

The duck.

Speaker 1 [28:18]

That's one. Do you want a card?

Speaker 3 [28:20]

In their card take one you have to read it. Yes Can't read this

Speaker 1 [28:28]

Speak about your topic like it's a dangerous secret you're revealing. Ah.

Speaker 3 [28:32]

I see. Well, I invite you to come along on this journey with me. With great power comes great responsibility. But fear not, I will guide you through it.

Speaker 5 [28:43]

through it all right so let's see

Speaker 1 [28:47]

All right, you have five minutes. Go ahead. What's the word, by the way? What's the title of the talk? German Open Talk.

Speaker 3 [28:54]

Well, I'm introducing European summer clothes.

Speaker 1 [28:56]

So are you happy if I choose open source, that's the word? Ah. Or German?

Speaker 3 [29:01]

Open source, let's choose open source.

Speaker 1 [29:02]

All right, open source in German, you can do quack. Can you do that for me?

Speaker 3 [29:05]

open source open source open source open source you don't understand

Speaker 1 [29:13]

You, yeah, you're overusing your powers, yeah.

Speaker 3 [29:18]

supposed to do that? Here's a dangerous secret.

Speaker 1 [29:20]

You have five minutes, you can do what you like.

Speaker 3 [29:23]

meet new friends, work on exciting open-source and AI, example from the SK Time code base in Python. Have fun with flags and join a program bringing together diverse set of stakeholders in industry, public sector, academia and open-source communities. If yes, then one, two, pep eight, Import this, the European Summer of Code, coming to you, well, last month, but you can still apply. Open source and open science for global societal benefits. The first European open source program, well done. With stipends for contributors new to open source, matching private and public sector projects with communities of open source communities, to be precise, applied projects with open technology, open source and research focus. The second batch of applications is still open until April 25, third batch released 26. We have currently three hubs in Europe, searching for more, Probable, the scikit-learn company, German Center for Open Source AI, that's us, and the Bruno Kessler Foundation in Italy for junior developers, open source communities and contributors, and applied projects. So we're inviting industry and public sector stakeholders also to propose AI projects that can be tackled with open source.

Speaker 1 [30:58]

Well done, thank you. Go to...

Speaker 3 [31:00]

ESOC DE, a dev, sorry, that's dev, should be, there's a letter missing, if you're looking for an open source stipend, see, apply for one of the projects, you can find them on the web page, your private or public sector, submit a sponsor interest form with a project, I know we can develop it with you, are you an open source project and community, join one of the hubs in Europe, or if you want to become a hub, contact us directly. Now, for open source projects and introducing the German Center for Open Source AI, the German non-profit umbrella with the largest open source AI footprint, working towards societal recognition of open source work, empowering independent open governance model for open source, the nascent expanding umbrella providing admin capacity to open source communities, events to connect with industry and application opportunities with open source, industry, public sector, a number of sponsored projects, we are growing, we are incorporated in the Schengen area, and some affiliated projects that are collaborating with us on the European Summer of Code, also open source projects, well known on here, we offer developer support, fellowships, work towards visibility and recognition of open source, help you with communications, okay, it's getting tired, right, should I not try to avoid the word, applying for grants revenue streams, physical hosting without overheads, no irreversible handover of assets, shared admin legal security, the poster and governance. We help you professionalize your open source community. We also offer the software ecosystem co-development shared maintenance opportunities, help with integrated research projects, join the German Center for open source AI. Who Who finds this exciting? Show of hands. Open source. Thank you.

Speaker 1 [33:14]

So while we get Nico on stage, we pick the next speaker.

Speaker 2 [33:17]

Can I keep the duck?

Speaker 1 [33:18]

No, we need that but I can I can give you afterwards but can I keep the

Speaker 3 [33:22]

Can I keep the power?

Speaker 1 [33:24]

Still no

Speaker 3 [33:28]

Respecto sponsorum.

Speaker 1 [33:35]

this is actually a great segue on we have chose the Harry Potter team for this year lighting talks and you're gonna make sense much more tomorrow so be here tomorrow Nico while you're setting things up have you chose your dares sorry have you chose one of your dares

Speaker 4 [33:54]

I'm already excited standing on a stage in front of many.

Speaker 1 [33:57]

Yeah, yeah, blah, blah, blah, right. Oh, yeah, yeah.

Speaker 4 [34:00]

Yeah, yeah

Speaker 1 [34:03]

It doesn't work like that. Do you want a card or not? Yeah. Brilliant.

Speaker 4 [34:07]

I'll take one.

Speaker 1 [34:09]

Well, you should pick two.

Speaker 4 [34:12]

Not the one you're like no

Speaker 1 [34:14]

No, I was not handing anything. I promise I don't know the cards, I promise.

Speaker 4 [34:18]

So that's the fun side.

Speaker 1 [34:20]

yeah that's the wrong side just one as you like you can change one if you like only one though

Speaker 2 [34:29]

I have to do too.

Speaker 1 [34:31]

If you want, yeah. Those would be interesting, yeah. Yeah, you can do those. It's fine. Come on, you're great.

Speaker 4 [34:39]

I don't have slides, so I take both.

Speaker 1 [34:41]

You don't have slides? All right, well... Kind of.

Speaker 4 [34:45]

to find a slide, you know.

Speaker 1 [34:47]

Yeah, it's up to you.

Speaker 2 [34:50]

I just

Speaker 1 [34:51]

I just want to say, since we need to have the best talk for the session, I dub all of you sitting in the front row to be the jury. So everybody here in the front row is going to vote at the end who's going to be the best talk, right? So it's on you then, right?

Speaker 2 [35:10]

Next talk is Luca

Speaker 1 [35:13]

You can vote for yourself, but that would be a shame though, right? But if you have powers...

Speaker 2 [35:20]

Afterwards, we'll be Luca with...

Speaker 1 [35:21]

Luca oh Luca oh Luca are you here I think you are so you brilliant okay get ready for the next one Italian cues that's amazing I know this talk brilliant all right take it away oh we have to say the challenge

Speaker 4 [35:36]

Sorry, forgot that. So I have to explain one slide using only words of five letters or fewer. And I have to tell the audience that my next slide is my favorite, even if it clearly isn't. All right, take it away. Interesting. All right, so that's going to start. The next slide is my favorite, actually. Niko? Dayla?

Speaker 1 [36:10]

Bro?

Speaker 4 [36:11]

North and I'm going to tell you a little bit about oh time is running really fast about like dependency management some things I experienced over the years and some really new stuff that I saw today actually as I'm following some release notes you should all do that on github or wherever your favorite open source package is hosted it worked somewhere and my notes are getting away so I think like probably all at some point just had some requirements file like this just said okay just install it into my environment if you have an environment or using like environment managers and they're the problem start when you use that in production we did that quite a lot of years ago when I was a working student and some point like everything together broke because like the versions did did jump and time is running really fast so I go a little bit faster and so you you will start to use like compilations using like pip compile poetry whatsoever there are different tools out there for that that you can do to like pin your versions and I we started doing that quite early I think most of you are doing that somehow probably because that really helps that comes thanks that comes with some risks you know like when you when you use like when you're like working with packages you and do like own packages so you're like doing Python packaging you are running into a security risk like dependency confusion text so let's say you have a module pan pie that you're using then somebody else may steal it kind of so that's why you may want to pin down the indices you're using etc and I think that's where pip compile started when I used that to run into issues and I started using UV as that came out at that time and that was the latest shit you know like I already had some experience with some linter with rough and that was really fast compared to all the other linters that available I think three four years back so then when you become came out you know it looks the same the the fire you're compiling there I think you can do life it it really is let's it works out you know it compiles that it writes it to the file nothing to worry about the the thing about that it's written in Rust. You may have attended some talks today that's usually way faster and I really was impressed by how fast it is. For us it took like 40 seconds to like 5 seconds of like this compilation time and that was really nice. So I am following them and quite intensely and I'm following those release notes and today I actually noticed okay I do have some like notifications like those you see here and this was released yesterday night at 2 a.m. or something probably some other time zone and and this is quite interesting I thought because you know like okay there's some new stuff coming up some like you see some pet and those are usually interesting and I was wondering okay what is this about and there's some new like tomlify you know like everybody knows yaml but why are those Python people using toml that's the question I'm still like trying to understand the differences they are both quite opinionated and there's now another one besides the PI project and that is quite cool because that tries to unify how different tools you know like poetry, pip, pdm, not pip but all those are like handling log files and UV as well and that one I think when you look at the actual enhancement proposal looks quite cool I think my time is running up I only have 30 seconds left so let me jump to some interesting things how does this look like now when I when I compile it so actually I have those packages and the interesting thing is I do have the wheel specified including the hashes so whatever tool I used to then install these things that is that can use the hashes to like verify that's really working out it can directly use this link to download the So it doesn't need to resolve anything and I think that's a really cool thing and just to close on that one That was the other one and I think that's really interesting to see like this discussions on peps so all please check out like Python and handman enhancement proposals Really interesting discussions on there

Speaker 1 [41:15]

Okay, why Luca comes in stage and you can pick your dares We can select the next speaker. Yes Let's see how many we can fit in today All right, we have Johan

Speaker 2 [41:30]

By Johan

Speaker 1 [41:31]

Experimenting tracking and LLM tracing using basic Python modules. Are you here? Brilliant, okay, yeah, I remember you. All right, did you want your dares? Shall we bring the box closer maybe? Well, we have glasses, I know they normally fit you very well from experience. I remember some fancy glasses you had.

Speaker 7 [42:01]

Let's do it with style then.

Speaker 1 [42:02]

Alright, yes. Okay. Wow. These are powerful glasses. Yes. Okay.

Speaker 7 [42:07]

I think I see, like, from the bottom.

Speaker 1 [42:09]

All right, do you want a challenge?

Speaker 7 [42:10]

Absolutely. Of course.

Speaker 1 [42:12]

How many do you want?

Speaker 7 [42:13]

I'm a maniac.

Speaker 1 [42:15]

I think I get it. You can be bold.

Speaker 7 [42:17]

I think I'll start with one, I'll read one, and then I'll tell you. All right. Okay. I'll pick for you. Wow. Name a totally fake Python package name I recommend seriously for 10 seconds. Okay. Actually... This is going to be cool.

Speaker 1 [42:30]

This is fitting you very well, actually.

Speaker 7 [42:30]

This is... Well, actually.

Speaker 1 [42:31]

Absolutely.

Speaker 7 [42:31]

Absolutely, actually.

Speaker 1 [42:33]

Yes, absolutely. Okay, the more you have, the more chances you have to win. No kidding.

Speaker 7 [42:38]

No kidding. Okay, oh, let me grab the whole thing. Okay, let's take this one. Introduce yourself like you had fair titles, such as extreme supreme debugger of the north. Yes. Yes? Unironically. Okay. Okay, go ahead.

Speaker 1 [42:55]

Five minutes of your time.

Speaker 7 [42:55]

Okay. Okay, good evening everybody!

Speaker 1 [42:56]

Okay.

Speaker 7 [42:59]

Okay, so my name is Luca and I am an, I don't know, machine learning burper at Xtreme, though I do not have any whatsoever kind of computer science education. This is why I've been delving deeper into the depth of the studies of data algorithms and data structures. How many leads code have I passed so far? Zero. But that's not a problem. I need to thank very much Alessandro Re, who is also a PhD, an MSc, a BSc, a DM in Dungeon Masters, to be precise, and also, I think, is a CA. And that's not me at all. That's the guy who taught me how to do this properly. I'm actually Luca. I'm not a PhD. I'm not a Dungeon Master. I'm a top laner in League of Legends, if you're interested into that. And just a Bachelor of Philosophy. So yeah. The beauty about this is there is something that I did implement in a Python library, and it's an implementation of a specific type of queue, which is the Italian queue. A queue or a line is a linear data structure, right, to manage data that should be processed sequentially. Well, an Italian queue is a slight twist, right, because it follows a certain kind of structure that is called a CISO strategy, something in, something out, or first come,

Speaker 1 [44:15]

That is true, I promise.

Speaker 7 [44:24]

Also known as first come, eventually served. And usually I would be explaining the theory behind it, but since I had so much time, I built a library, a Python library to do this. So we're going to spin up the best editor that you can have available. Neovin, of course, not Cursor, I'm sorry. I'm sorry for the JetBrains people out here. I also stole so many of your stickers today. But still, can you see? Okay, that's another point on my bingo card. So let's zoom this, and there we go. You can also use the blazingly fast package manager to install this amazing library from directly in my GitHub repository, which can exist actually right now. And the usage is fairly simple. You will also be encouraged to naturally use wireless operators and the newest features to enhance your Python style guide. And as you see, we can just run this like this. Oh, I need to divide this. The implementation is fairly simple, it requires a node module, oops, so many errors. The node is a structure that has a left and a right, and it also implements a recursive insert method, which is private. And this is actually using a recursive function because I'm too cool for loops. The remainder of the structure is, oops, not this one, is here, there we go, it's the Italian Q on itself. There are some pseudo-implementation left to the reader, like here, but the gist is this one. The init and push methods are fairly easy, it's the pop where the magic lies, right?

Speaker 1 [46:01]

Oh, bingo. Bingo. Thank you, mate.

Speaker 7 [46:02]

You're welcome.

Speaker 1 [46:03]

You're welcome.

Speaker 7 [46:04]

We can share the prize for that. And so you can see that if basically the whole of the structure is the pop method, and if you're the next in queue, we extract the successor, we update the queue, and if there is somebody on the left or on the right, they will get priority. In the end, we're still returning the person who's being served. That's great. OK. So this is as far as the implementation with bugs, right? That's OK. But it's part of the nature. Or maybe that's a feature. But let's just see it in practice, right, when it does work. You see the counter, person A gets in, now there's a spot on the left, so B comes in. There's a spot on the right, so C comes in, and now it's a turn for D, and a turn for E, and this naturally follows where person F will join there. If we do a proper order of arrival, Alessia, Bob, Carla, Dean, and Benevan will come. Bob, Carla, Alessia, and Dean, and then Alessia again will be served.

Speaker 1 [46:56]

It's fairly easy works

Speaker 7 [46:58]

It works perfectly, and it's, of course, linear. The complexity, and this is where all the codes will appear and become of use. Given an element pushes, of course, all of one. Pop is all of nine three-fourths. Space complexity is, of course, all of n plus one. Mind complexity, well, that's a whole story. Not to think about time complexity, well, you have to deal with this, right? And I dare you, you shouldn't be joking this seriously about how I'm doing this because there you say, okay, this is just a Latin talk, Luca, right? Well, I dare to say there is so much evidence for this. Thank you.

Speaker 1 [47:54]

Thank you very much. Thank you. That was amazing. Thank you guys. Give the glasses back. Thank you. Okay. I know you're Italian, you know That's your jacket Can we have one more stage please? Do you want dares? Do you want any of those?

Speaker 2 [48:14]

Maybe this is a good

Speaker 1 [48:17]

Let's go for challenges then.

Speaker 2 [48:22]

Can I try that one I need to touch my laptop it always

Speaker 1 [48:31]

It's over there, the power outlet. Don? Yeah. Somewhere? Oh, someone is coming. Oh, my God, that's very long. Sudarsan? Thank you. Surendral Hal. Apologies for butchering your name. Are you remotely here? Are you here? The title is, When the Law of Large Numbers Fails, how to generate hyperlocal geo data sets from panel data okay I'm gonna nobody's here all right pick another one we save it for tomorrow please delete your unit tests we don't have a name is somebody here all right okay brilliant get ready to get on stage okay fantastic so pick one card or as many as you like don't get stressed out it's gonna be fun

Speaker 2 [49:35]

Thank you.

Speaker 1 [49:36]

No, that's the other side you have to turn the cards, right?

Speaker 2 [49:39]

Bow to the audience before or after delivering your most important point like that

Speaker 1 [49:45]

like that all right yes that's a test

Speaker 2 [49:52]

I wanted to start with one most important sentence, and maybe then you can already forget about the whole presentation, hopefully. So the TIDR is, you need to have LLM observability, you need to check the quality of your LLMs and how it's going. And there are all sorts of platforms out there, cloud platforms, cloud providers, complicated frameworks. You can do it with only normal Python standard logging. That's the basic method standard import logging with that You can resolve everything also locally as I'll gonna show Yes So and if you want to Get the more extended version of that. You might even want to listen a little bit more so LM observability. Um So typically everybody knows LLMs are hard to debug. We also have these compound systems as we have heard previously. Compound systems are agents. So you have tools that you provide, you have very specific prompts, you have a rack system, vector based search, everything you throw into one LLM contextualized.

Speaker 4 [51:09]

Soup.

Speaker 2 [51:10]

And then it magically gives hopefully some good-ish kind of output. And it's hard to debug in the end. You need to trace down where the problems occur, when problems occur. So you need to be able to trace your LLMs. And also you want to be able to debug further, also see which actions are costly, which actions take time, et cetera. So traceability, observability, monitoring, debugging, et cetera. and so currently there are these in cloud providers that offer solutions for all of this we have one premium premium sponsor here they're called a lock fire and they also have that as a business model and also a length Smith from link chain and then we also have a helicon arise many many many more make money with these platforms and your cool GIs and web servers and don't know what and many advantages so it's very feature-rich has built-in analytics and it's a complicated vendor login external data dependencies external Python modules blah blah blah so I propose struct lock but you can also use normal standard Python logging import logging works also with a few lines of code more So for the demo, I just show you struct log one time It's structured logging structured logging just means You dump JSON into the logger and it saves it to a JSON JSON L format Text file and then afterwards you query that stuff with duck DB or with whatever you want to query and you get out very nice tracing very nice statistics and just the same way and That's what I want to show you. So I give you one example. That's just basic stats and the other example is multi-hop sequences with um i don't know conditional several steps things like these um lm workflows so quick example how many time is left one minute 14. oh wow um so uh basically i just i just show you okay you have to somehow believe me so what what I'm doing here so I'm running for example the whatever it's not the LLM.py it's the LLM dot yeah and it's a simple LM oh oh and it's the run and this runs the simple LM script it's just basically always the same prompt with different questions runs through it in the end I make a simple analysis that just is basically lots of queries on a JSON L file so here we have the JSON L file that has been created from that run and I run afterwards the simple analysis and it gives me some neat looking tables gives me some statistics it gives me the number of successful requests, average duration, average number of input, output tokens, etc. That one is easy. That's just standard debugging statistics. I think everybody knows that. But then also you have the advanced use case and that is advanced LM. That takes much more time to process. What happens here is lots of prompting, conditional. You can borrow it, right? And... Thanks so much. Yeah. You have to believe me that you see these cool nested dependent prompts which are conditional also, which look exactly like in the UIs.

Speaker 1 [55:25]

Fabulous, thank you. Do you need anything? I didn't catch your name. Ziu, right, right. Come on stage. Are we... Yeah, I know he's been speaking already. Thank you. Sadly... Speaking already? Sadly, we have to make the last speaker today, which is going to be, I don't think, there's no name, remember to put your name in the proposal, setting beautiful tables with Python and tips, or types, do we have the gentleman or the lady in the room, okay, fantastic, alright, sadly this is going to be the last entry for the day, We're going to save some for tomorrow, of course, and more will come.

Speaker 5 [56:28]

That's okay.

Speaker 1 [56:30]

All right Okay, you can take your glasses over there give you the funny bunny ears enjoy You are for five minutes of time unless you want a challenge as well

Speaker 5 [56:45]

Give me a challenge, all right

Speaker 1 [56:51]

What was that, sir?

Speaker 5 [56:52]

refer to the slides like they were sentient. For example, let's see what slide 3 has to say about it.

Speaker 1 [57:02]

Take it away. It's gonna be fun

Speaker 5 [57:04]

Okay, so welcome to this talk. Do I have here some data engineering professionals, people that write tests? Raise your hand, come on, raise your hand. See, okay, there's a lot of you. Okay, cool, cool. So, opinions are not my own, okay? I exaggerated them for fun. So, I would like you to entertain with the idea of completely deleting your tests on ETL pipelines. Just don't do them. They're a waste of time, okay? So, let's remember what, and see what slide two has to say about it, let's reminisce about the good old times, where things were simple, right? This is probably what your ETL pipeline looked like. So, it was simple, it had three things, it had a simple technology, was that one, okay? It was not super complicated, you could make changes, you could refine them with time, and not spending a million years maintaining and running those fucking tests. You You probably know already where I want to go probably, but you know. So let's see what slide three says about what the landscape today looks like. So this is not my data pipeline, but you know, pretty close, about 50 elements, 86 technologies interacting, permissions, problems for which you never have access. So you know, a fun landscape indeed. Android 4 also tells us about the landscape, how easy it is to make a simple choice in building such an ETL pipeline. Come on. I mean, can't you just choose one? Go ahead. It's easy doing all of this, right? It's easy. So you can just pick one, then probably you need the other one. It will install a million dependencies, especially in Python. It's very fun handling all of this. So you will get a thousand packages. You can't even create an environment on your computer anymore. You need to deploy it anywhere else. So, you know, let's actually talk about those tests, knowing that all of this is very simple, okay? And I will bring up my favorite IDE, which is not a text editor. It's an actual IDE, and it's PyCharm. So just use an IDE, please. Thank you. So what do we see here? Well, this is your average. test, you're parameterizing a million things. And here, I'm testing a simple multiplication. I don't know if you see it. It's line 39. That's the test I'm writing for. I'm writing a test for that function. I need to reproduce my data frame exactly. I need the correct data types. I need to write the test. If I change the behavior, I need to look to this This fucking... Sorry.

Speaker 1 [59:49]

and

Speaker 5 [59:50]

To this very simple to read the parameterize that probably looks like something you've seen and I just wanna Quit and start a farm somewhere when I read this. Okay, so I Would basically yeah, no not not working any longer when I see any of this and so please For example, imagine that you have to migrate this to PySpark, right? Your workload is scaling up you need to do more things. You're gonna get more complicated. If this is your test, do you know what I think the best way of entering the situation is? This one. Just delete it. You don't need it. You don't need it for anything, absolutely anything. So if you have to take one thing away from this five-minute talk, which is actually a little bit like my opinion, but don't take me too too seriously. The actual bit you have to take home is remember to delete the unit tests.

Speaker 1 [61:02]

that was amazing thanks very much you forgot about the bunny's ears though but yeah that's fine all right okay let's have the last speaker of today and then afterwards we're gonna vote for the best talk of the day and we have all the proposals over there and then we go home all right take it away do you want any challenge

Speaker 8 [61:30]

Yeah, sure.

Speaker 1 [61:31]

Can you hand him the card, please?

Speaker 8 [61:44]

no deliver the next slides with absolute seriousness no matter what's actually on it good thing I prepared a serious talk

Speaker 1 [62:01]

I don't think, oh yeah, slides over.

Speaker 8 [62:07]

Just going to mirror it and that's nice Okay in all seriousness Who have you already crunched data with pandas? Who needed to actually export the resulting data frame in a human readable format like PDF or something Who used Lata to do it Who actually enjoyed the process? Yeah, me neither. So, the thing is there's a new kid on the block and it's called typest. It's like LaTeX, but it sucks less. And it compiles much faster. The only downside is that there is no data frame to typest method. At least there wasn't until I built it. So thing is pretty simple. You have your data frame, this really nice data frame. You do your stuff with it. And then you want a PDF. And what this package, it's a personal project of mine for your bingo cards. It's called Pipest. And what it does is it just takes your data frame and makes a table from it. And you can actually style it right in here, so if you want the lines to go there, if you want to have a zebra pattern on it, you can all do it in Python. And then you just tell it to render, write it to a types file, and then you compile it and be done with it. The output in types looks like this.

Speaker 2 [63:55]

and

Speaker 8 [63:56]

The title of my talk was actually setting beautiful tables and I noticed that I didn't bring any example. I Won't rectify this This is it

Speaker 1 [64:22]

all the speakers on stage please so we do the price delivery actually we have two prices one is of course the snake the second best talk gets the bunny here's so we don't have it for tomorrow or maybe we do I don't know all right so I apologies folks I don't remember all of your names so I just go in order Please take all your stage. It's all to you. You're on the spotlight nowadays. All right Okay going in order Please delete your unit tests people in the front row How many votes we got for Zeus is it? Z sorry, how many votes for for Ziv? anyone voting That's actually cool, all right one votes very cool very cool of you well done all right then we have I know you give me a sec Johan brilliant oh yeah that's your talk Johan ML experimenting tracing and NLM how many votes we got from for the guy zero yeah thank you thank you remember it's just like about delivery is not about the topic on anything it's just like you have fun what do Luca how many votes for Luca all right so far Lucas leading when we have friends how many votes for open source in Germany how many votes okay still Luca right I need to have the second place as well so how many votes do we get from France you can vote multiple times yes if we can't start over again okay you can vote all right so it's on me not explaining well apologies right by the way for those doing bingo already are we handing the prices tomorrow just give me the the the cards you had win with your name on it after the lightning talk session and we will hand in the prices tomorrow first thing all right let's start again please delete your unit tests how many votes all right for okay we got a full solid one Johan how many votes one okay for one okay good how many votes for Luca, three, four, five, six, seven, eight, nine, 10, 11. Okay, Luca's leading, brilliant. How many votes for France, Germany, open source? Three, four, five, six, seven, eight, nine. I think it's more than Luca, isn't it? How many was Luca? Okay, that's go second, okay, I like that. How many votes for, who do we have next? I don't have your name, sorry. What's your name? Yeah? Tillman. Tillman, okay. How many votes for Tillman? One, two, three, four, five, six, right? How many votes for Nico, UVMPEP, 751? One, two, three, okay. And last one for John, the very first and the most brave one of everybody. How many votes? All right, brilliant. I think we do have our winners. Luca and friends, please claim your prizes. Believe it or not, Italian queues do work that way. That's something I can promise you. All right, congratulations. Can anybody do pictures? And this is the first sets of lighting talk speaker for PyCon D2025, a huge round of applause to everyone. Oh yes, of course, I was almost forgetting, every speaker actually gets a prize, and what we have is a fantastic poster of the Zen of Python they can bring them back home Johannes is fetching it so they can show it to the audience it's coming all right well we have many all right yeah all right so everybody joining and this will happen tomorrow as well so this can be an incentive for people to hand in their proposals so they get a fantastic poster well done all right thank you very much for joining and I hope to see you tomorrow