EuroPython 2026 Recap

EuroPython just concluded in Krakow, Poland, and I was there as a member of the PyCharm team, manning a booth for three days, giving a talk on deployment, and more. This post is an image-filled recap of the full week.

Tuesday, July 14

I departed from Boston Monday evening and arrived, via Munich, in Krakow on Tuesday afternoon. Krakow and indeed Poland itself was a first for me, and I can report that the location was lovely. The conference venue (seen here on the right in the photo) was right on the river, and there was even a small amusement park and hot air balloon nearby. The Speakers’ Dinner on Tuesday evening was just a short walk from the venue.

Across the street from the venue is the Wawel Royal Castle, which is mostly open, meaning you can walk around it freely.

The Speakers’ Dinner was on top of a local building along the river, and downstairs on the grass there was a huge screen setup to show the France vs Spain World Cup semifinal game. I didn’t manage a picture of the screen itself, but here is the view looking down from the venue.

Despite my jet lag, it was a lot of fun to see familiar and new faces at the event. By accident, I walked over from a nearby hotel with none other than Guido van Rossum himself. Both of us were required to show a speakers’ dinner QR code at the door, which I found amusing.

At the venue, I ended up having a good conversation on the balcony with Katie Bickford and Imogen Wright, who didn’t mention they were giving the opening keynote the next morning!

I ended up leaving after two hours, sooner than I would have liked, but having been up for well over 24 hours at this point, my bed was calling.

Wednesday, July 15

PyCharm had a full team at the conference. In this picture are four Developer Advocates (myself included), our Head of Python Ecosystem, PyCharm’s Product Manager, and a Technical Content Creator who is, at this writing, now an AI Developer Advocate.

We had demos to give, swag to give away–sunglasses, tote bags, and a raffle for a longboard–as well as a beading station that proved quite popular. PyCharm continues to add new features, especially around typing, and ever-more AI options for agentic workflows, supporting basically every model out there.

Also, PyCharm 2026.2 came out today, so you can read all about it on the official blog post and video if you are interested.

Back to the conference, things started off with an opening discussion, ably led in part by my colleague Jodie Burchell (seen here).

I didn’t manage to take a picture of Imogen’s incredible keynote, but I highly recommend you give it a watch once all videos are up on the EuroPython YouTube channel.

During the day, traffic to our booth and those of other sponsors was robust.

I was pleased that Django had its own community booth downstairs, showcasing the new 6.1 release in a few weeks.

And Paolo Melchiorre, one of the organizers of Django on the Med, development sprints for Django, was tirelessly promoting the event.

There were talks all day long, although aside from the opening keynote, I was mostly at the PyCharm booth.

Thursday, July 16

Thursday morning was another day of talks. You can see just one of the beautiful rooms here, in this photo.

My own talk was in the late morning, on Python Deployment, and I was very pleased–and somewhat surprised–that the room was completely full well before I started.

There was also a lively q&a afterward, as it turns out deployment is still an interesting topic to everyone. The official video of the talk will be up in a few weeks.

EuroPython serves as a large umbrella for many other Python-related groups to meet and host events, most notably PyLadies, which had a lunch right after my talk. I was invited to attend. My colleague Jodie was MCing the event.

And Kristel from the PyCharm also stood up and spoke.

At the end of the day, there was a EuroPython Social Event that was well-attended, but I had had a full day and opted instead for a quiet dinner and stroll along the river. It is light until after 9pm in the summer, and the riverbank is full of people sitting, eating, having a drink, and generally enjoying the pleasant weather.

Friday, July 17

Friday morning I managed a 6am run with my ultramarathon colleague, Galina. Then it was off to the PyCharm booth again at 8am and a final day of speaking with attendees and sneaking off to a few more talks, when possible.

Paul Everitt from PyCharm gave the closing keynote, which I, again, highly recommend watching when it is available online. It was about the Python community, AI, and really all the current things shaking up our ecosystem.

In the evening, I had dinner with Paolo and Galina, finding a really nice steak restaurant in the old town.

After that, it was drinks with many people and recapping the full event.

Saturday, July 18

My flight was in the early afternoon, which meant I could wake up when I felt like it (turns out that’s 7am), go for a run, have breakfast, pack up, and share a cab with a colleague to the airport. It was a long way back flying against the jet stream via Frankfurt, but I made it home in the evening Boston time.

Takeaways

My quick takeaway from the event was that whereas a year ago the discussions were around AI, which models to use, and whether it was “a thing,” this year it is clear to everyone that AI is here to stay in one form or another. Agentic workflows are all the rage and the focus is less on a specific model and more on “the harness” that feeds information to it and prompts it to try again and again until some measure of success is achieved.

There are built-in agents available like Claude Code and Codex, and more complex options via OpenClaw, but developers at larger companies all reported quite complex internal tooling. The gulf I see emerging is between developers using off-the-shelf tools and those in enterprises who are fully committed to agentic workflows and, presumably, are seeing better results.

The tricky thing with AI is that there are still no truly objective ways to measure performance. Costs can be measured and are increasing. But actual performance is much more difficult. Yes, there are benchmarks, and everyone tries to craft their own evals–though good luck finding a common definition of what “evals” really means–but really, it’s all based on feel and vibes, which are unsettling metrics for most programmers.

PyCharm continues its mission to support as many workflows as possible. That means virtually any model or agent you want to use is supported. The IDE itself is, to me, more important than ever if a developer wishes to add value as a human-in-the-loop here, but it remains to be seen how much developers will even read most code, let alone write it.

I have my own reservations about a full shift towards agentic-all-the-things while relying on integration tests, but it is clear that is where things are headed at the moment in many companies. At the same time, senior Python engineers are urging caution, pointing out that fundamentals and human judgment still do matter, maybe more than ever, despite this shift towards handing more and more of the workload over to AI.

I’m curious to see what kind of conversations happen next year at this event, which will mostly again be in Krakow.