Reflections on Current 2025: Streaming, AI, Beignets, and The Buffalo Bills?

As the mental (and possibly physical) hangover begins to wear off, I figured I’d take a moment to reflect on my recent trip to Confluent’s 2025 Current conference, conveniently held right here in my home state of Louisiana.

It was a week filled with brilliant people, incredible sessions, and more powdered sugar than my bloodstream was prepared for. And honestly, there’s no better setting for a conference about streaming systems than a city literally built around one of the world’s most iconic streams: the Mississippi River.

New Orleans has always been a place defined by rhythm, resilience, and flow — three words that also describe a healthy data pipeline. From the steady pulse of jazz to the unpredictable currents of the river, it’s a city that understands how to live with motion instead of fighting it. Which, come to think of it, is exactly what we do in distributed systems.

AI and Streaming on a Collision Course

If there was one recurring theme throughout the week, it was the growing convergence of AI and streaming. The two were inseparable, from the keynotes and sessions to hallway conversations and evening chats over cocktails.

As folks who work in streaming systems, we’ve spent years trying to perfect how to move data efficiently through a system. Many of those same principles are now being applied to teaching it to think while it moves. The coupling of real-time decisioning with real-time intelligence is becoming easier, more powerful, and more tangible every day.

The keynote announcements around Confluent Intelligence and countless other improvements towards converging AI and streaming make long-time pipe dreams feel within reach. It’s all exciting, maybe a little terrifying, but seemingly inevitable.

The Mississippi doesn’t ask permission to change course, and neither does innovation.

People Who Made It Memorable

While AI was clearly the star of the show, it wasn’t the only highlight. The people made it special.

• Adi Poluk gave what might’ve been my favorite analogy of the week: tying streaming to the unfolding of history, complete with a little Mardi Gras flair. Clever, visual, and the kind of metaphor that makes you rethink how you explain what we do.
• Tim Berglund continues to prove that if he ever leaves the streaming world, late-night television will have a new contender. The man can turn Kafka into punchlines and monologue with the best of them.
• The Q&A sessions were some of my favorites, especially as an open-source contributor. It was great to meet and interact in person with folks I’ve known online for years, like David Anderson and Matthias Sax.
• And of course, Batchy made sure everyone knew it when he was around. From crashing Tim’s late-night show to leading a full-fledged New Orleans second line, he was everywhere. My kids were thrilled with the mini versions of him I brought home.

While those were many of the Confluent-led highlights (sans the Buffalo Bills incident), the real magic happened in the conference hall — the personal, technical, and wonderfully human conversations.

A few other highlights (but by no means an exhaustive list):

• Finally meeting ShadowTraffic creator and solopreneur Michael Drogalis, the man powering the keynote (and who knows how many others) and providing a never-ending stream of insights online (follow him).
• A long chat with author and serialization enthusiast Scott Haines from Buf about Protobuf, engineering culture, championing change, and yes — Legos.
• Extended discussions with Derek Troy-West and the Factor House team on observability, governance, lineage — all those intangibles that turn good systems into great ones.
• Talking Kafka, Pulsar, logo design, and plushie quality with Christine Schaefer and the crew from StreamNative.
• The impressive memory of some of the minds from the wonderful team at StarTree, folks that power Apache Pinot for not only recognizing me on the floor, but also remembering every member of my team they've ever interacted with.
• A great discussion with Joey Waldman from SuperStream about continual, AI-driven improvements for streaming systems.

A City, A Conference, and a Community

Only in New Orleans could an AI-driven crowd project somehow choose the Buffalo Bills over the Saints. I get it, Josh Allen’s great, but in Louisiana? That’s rough.

And only in Louisiana could a tech conference include a Grammy Award–winning zydeco band taking the stage — and somehow have half the crowd still talking about stateful operators while indulging in classic Hurricanes (possibly building up the confidence to learn to two-step). It was a surreal, wonderful reminder that this city doesn’t separate work and rhythm; it just blends them until you can’t tell the difference.

In true local fashion, I also found myself confronting one of the classic Louisiana culinary controversies: gumbo allegiance. Growing up on the other side of the state, I’ve long been skeptical of New Orleans-style gumbos. It’s basically a Crips vs. Bloods situation here — okra vs. no okra, tomatoes vs. no tomatoes, etc. (It was solid, I'll give them a pass – but I'll give the edge to the other side of the state).

But the thing that really stayed with me wasn’t the food, or the music, or even the sessions. It was the community. There’s something special about seeing so many people — engineers, architects, and tinkerers — all passionate about the same evolving space. The hallway chats, the “oh, you’re that GitHub handle” moments, the late-night serializer debates — that’s the stuff you can’t stream.

Where The Current Leads Next

I left New Orleans a little wiser, a little sleep-deprived, and a lot more caffeinated, but also energized (despite the real-world backpressure of Baton Rouge traffic that nearly doubled my trek home).

The intersection of AI and streaming doesn’t feel theoretical anymore. It’s here, now, and moving fast. The Mississippi doesn’t stop when it hits the city; it keeps carving new paths, finding new ways forward. As forward-thinking engineers, we have to do the same as we navigate this AI-driven riverboat down the stream of data.

We design our levees, build our checkpoints, monitor our flow, and the current still finds ways to surprise us. That’s what makes this work fun.

Here’s to whatever bends the river takes next—and to all of us just trying to keep the stream steady. So until next time Current – Laissez les bons temps rouler!