Current is about to kick off in New Orleans, the city where French culture, live jazz, and late-night energy meet. The timing couldn’t be better: Halloween week, when Frenchmen Street turns into a wild parade of costumes, brass bands, and chaos. Can’t wait.

This year’s agenda can be split into two clear themes:

• Apache Flink leads with almost a fifth of all sessions, covering performance tuning, SQL features, and deeper AI integrations.

• AI and agent systems form the other big theme, with talks on multi-agent orchestration, event-driven AI pipelines, and real-time GenAI architectures. That’s where the ecosystem seems to move: from data in motion to decisions in motion.

The rest dives deep into Kafka’s core: KRaft, rebalance protocols, the Kafka protocol itself, and Iceberg, Tableflow, and governance topics that show how streaming is fusing with the lakehouse world.

Obviously I’m excited to hear what Jay and his team will announce at the keynotes. Current has become the moment of truth for the streaming world, the place where you feel the pulse of the ecosystem (on top of meeting all your streaming friends!). It’s where the "current" state and future direction of streaming are defined.

Below are five sessions I’m most excited about. Each of them captures a key direction the streaming world is taking this year: AI meeting Flink, governance meeting lakehouses, and enterprises finally treating data in motion as a core product. It's going to be exciting!

StreamLink: Real-Time Data Ingestion at OpenAI Scale

Adam Richardson, OpenAI

How does a leading AI company architect its streaming infrastructure? This talk is a rare chance to peek behind the curtain.

I remember talking with Jigar Bhati, one of OpenAI’s technical staff, right after their Kafka Summit London session on simplifying Kafka consumption for their teams. We went deep into how they govern internal Kafka usage, why proxies are key to balancing flexibility and control, and how onboarding works at that scale.

StreamLink feels like the logical next step: moving from managing users to managing the data itself. Seeing OpenAI connect Flink, Iceberg, and Kafka to power AI workflows will be a masterclass in how data engineering evolves when AI becomes the customer. Forty-five minutes of pure learning.

The Kafka Protocol Deconstructed: A Live-Coded Deep Dive

Mateo Rojas, LittleHorse

This one is for anyone who truly enjoys understanding how things work under the hood. Mateo Rojas from LittleHorse will literally rebuild a Kafka broker from scratch—starting with a raw TCP socket—and walk through every layer: wire protocol, batching, replication, offsets, ISR, idempotence, everything.

I’ve been following Colt (CEO) and the LittleHorse team for a few years now. Their work around durable execution and workflow orchestration fascinates me. The premise is simple but powerful: if your app is critical, it should be distributed and resilient by default—and Kafka is the perfect backbone for that. LittleHorse competes in the same space as Temporal, Restate, and Golem, but they’re building it with a unique vision and a great aesthetic (yes, I love their logo). Seeing them go deep on the Kafka protocol itself connects everything—durable workflows, streaming, and the core foundations that make this entire ecosystem possible.

The Evolution of Notion’s Event Logging Stack

Adam Hudson, Notion

I’m a very early adapter of Notion, it's where I put my whole life. Hearing how they manage (my!) data at scale is an obvious "I need to go". They process billions of events every day, using Kafka, Snowpipe Streaming, and Apache Pinot for real-time analytics.

I’m excited to reconcile my #1 software with my #1 technology, and see how they’re working with it at their scale. I’m genuinely curious about the kind of data they move, the tradeoffs they face, and how they keep the product fast and reliable for millions of users like me.

Diskless but with Disks, Leaderless but with Leaders: A KIP-1163 Deep Dive

Greg Harris, Aiven

This session dives into one of the boldest architectural proposals in Kafka’s history: KIP-1163, which introduces the concept of Diskless Topics. It rethinks how Kafka stores and replicates data in the age of hyperscale cloud. The promise is simple: lower cost, higher flexibility. Moving away from traditional broker storage reshapes how Kafka delivers durability, availability, and efficiency. This talk will explore the tradeoffs and what it means for how we’ll operate Kafka clusters in the coming years.

I admire how Aiven is pushing what was a status quo in streaming. I remember chatting with Filip Yonov (Head of Streaming) during one of our webinars, about where the ecosystem was heading, cost, governance, and automation kept coming back. This session feels like a direct answer to that conversation: how far can Kafka be re-engineered at its core to fully embrace the modern cloud-native paradigm?

Future of Streaming: Emerging Trends for Event-Driven Architectures

Matthew Walker, JP Morgan Chase

When a bank the size of JPMorgan Chase (over 300,000 employees) starts talking about event-driven architectures, you pay attention. Matthew will outline how they’re modernizing decades of mainframe data systems into a next-generation streaming platform built around Kafka. It’s the story of a financial giant moving toward an automated, data-driven backbone.

For me, this is extraordinary. We often talk about streaming in the context of tech companies, but seeing it reshape one of the most traditional industries in the world changes the perspective. I’m curious to see how they make streaming work within the rigor and regulation of global finance. Hearing their vision for what comes next, where compliance, AI, and real-time data meet, feels like getting a front-row seat to the future of enterprise data.

Playing Chess by Scaling Multi-Agent LLM Play with Kafka Queues

Stéphane Derosiaux, Conduktor

Lastly, a more relaxed talk. I'll try to demo Multiverse Chess, where different opponents compete in parallel against one player. It’s an example of how streaming platforms will handle complex AI orchestration and their coordination challenges that come with agentic systems.

Chess was my world growing up. I competed for years. This talk is where this passion meets AI. It should (will) be fun and obviously chaotic, but also shows something serious: that the next generation of AI systems will rely on streaming to operate at scale.