Hello, Agent!: The podcast at the intersection of data & agents

Building interruption-tolerant agents with durable execution with Dominik Tornow at Resonate HQ

Dominik Tornow [https://www.linkedin.com/in/dtornow/], Founder and CEO at Resonate HQ [http://www.linkedin.com/company/resonatehqio/], discusses simplifying distributed systems and why durable execution is becoming critical in the age of AI agents. We explore how agents introduce new challenges for distributed systems, why long-running autonomous workflows require interruption-tolerant execution, and how durable promises provide a simpler mental model for building reliable multi-agent systems. Key Takeaways: 00:00 Introduction. 02:20 A career start during the rise of AWS and cloud computing shaped Dominik’s focus on distributed systems. 04:10 Separating the concept of durable execution from specific implementations led to the idea of interruption-tolerant systems. 05:15 Failures as interruptions — voluntary or involuntary — lead to a broader, more practical model for resilient execution. 12:40 Simplifying complex concepts and why engineers often struggle to align their mental models when designing distributed systems. 14:10 Dominik defines an agent as a combination of a model, tools, and a system prompt, while distinguishing between an agent definition and a running agent instance. 23:30 Many effective agents are simply tool-calling loops, despite the complexity of the surrounding ecosystem. 26:20 Running agents in enterprise environments introduces distributed systems challenges, especially when workflows involve long-running or expensive operations. 33:15 Unlike traditional microservices, retrying agent operations can be extremely expensive when LLM calls or data pipelines are involved. 40:20 A durable promise is a simple mechanism for coordination and recovery in distributed agent workflows. Resources Mentioned: Dominik Tornow [https://www.linkedin.com/in/dtornow/] https://www.linkedin.com/in/dtornow/ Resonate HQ [http://www.linkedin.com/company/resonatehqio/] | LinkedIn http://www.linkedin.com/company/resonatehqio/ Resonate HQ [https://www.resonatehq.io/] | Website https://www.resonatehq.io/ "Think Distributed Systems" by Dominik Tornow [https://www.manning.com/books/think-distributed-systems] https://www.manning.com/books/think-distributed-systems Temporal (durable execution platform) [https://temporal.io] https://temporal.io Thanks for listening to “Hello Agent!: The podcast at the intersection of data & agents.” If you loved this episode, let us know with a 5-star review! Remember to subscribe so you don’t miss an episode. To learn more about Redpanda, visit redpanda.com #RealTimeData #DataStreaming #Redpanda

Designing secure architectures for agent-driven AI workflows with Nicolas Dupont at Cyborg

Building reliable agents begins with a strong data foundation. As AI systems move from experiments to production, teams need clearer ways to connect data, models, and real-world decision-making. In this inaugural episode of “Hello, Agent!” we talk with Nicolas Dupont [https://www.linkedin.com/in/cyborgnico/], Founder and CEO at Cyborg Inc. [https://www.linkedin.com/company/cyborginc/], about how agents are changing the way teams think about automation, streaming data, and system design. Nicolas shares practical lessons from building Cyborg, what breaks when agents scale, and how engineers can design systems that stay flexible as complexity grows. Key Takeaways: 00:00 Introduction. 02:00 Nicolas’ focus on solving technical problems for the real world. 05:43 Programming in C builds a deep appreciation for low-level systems. 09:39 Securing proprietary data matters more than models in enterprise AI. 15:32 Vector data can expose sensitive information to attackers. 20:59 Auth systems are often weaker than expected, even for sensitive data. 25:03 Agentic AI removes the perimeter, making zero trust essential. 30:06 Poisoned data is hard to trace across agent systems. 35:07 Deleting keys makes sensitive data unreadable immediately. 40:48 Modern cryptography is no longer a bottleneck, making strong encryption practical. 45:06 Zero trust isn’t optional — the future of agentic AI depends on it. Resources Mentioned: Nicolas Dupont [https://www.linkedin.com/in/cyborgnico/] https://www.linkedin.com/in/cyborgnico/ Cyborg Inc. [https://www.linkedin.com/company/cyborginc/] | LinkedIn https://www.linkedin.com/company/cyborginc/ Cyborg [https://www.cyborg.co/] Inc. [https://www.cyborg.co/] | Website https://www.cyborg.co/ Cyborg Blog: Vector Embeddings Are Not One-Way Hashes [https://www.cyborg.co/blog/vector-embeddings-are-not-one-way-hashes]  https://www.cyborg.co/blog/vector-embeddings-are-not-one-way-hashes Thanks for listening to “Hello Agent!: The podcast at the intersection of data & agents.” If you loved this episode, let us know with a 5-star review! Remember to subscribe so you don’t miss an episode. To learn more about Redpanda, visit redpanda.com #RealTimeData #DataStreaming #Redpanda