Context Engineering for the Agentic Era

​For the last few years, our job was to perfect the prompt. We built sophisticated RAG pipelines to guide the LLMs toward better, more factual answers.

​Now, our job is to build agents that act. Context Engineering has been elevated from a 'prompting trick' to the central discipline for building reliable, production-grade AI. The architecture is no longer a linear pipeline. Instead, the intelligence of the system now resides within a cyclical reasoning loop where the model must plan, use tools, and observe outcomes.

​This new agentic workload raises a new class of engineering questions. How do we build effective short and long-term memory for an agent? How do we design for tool selection as the number of available tools explodes? And what does our retrieval infrastructure look like when it must handle the relentless, programmatic workload of an agent instead of a human?

​This evening is for the builders who are facing the hard problems of making agents work. We're gathering Stockholm's AI community to discuss this new frontier.

​Speakers:

​Speakers:

​​Jo Kristian Bergum, CEO HORNET.dev

​​Lester Solbakken, co-founder HORNET.dev

​​Oscar Täckström, Chief Scientist Sana.ai 

​Program

18:00 Doors open & check-in 

​​​Grab a drink. Chat with other builders.

​​​18:30 The road from prompt engineering to context engineering 

​Context is a finite resource; reducing context rot and maximizing information relevance is key to enabling agents to navigate complex problem spaces effectively. As agents retrieve context from their tools, memory, and retrieval systems, context engineering has become the art of curating this high-signal information to guide the agent's desired behavior. In this talk, we will trace the evolution from engineering prompts for our language models to the challenges of managing context state for autonomous agents solving long-horizon tasks.

​​Presented by Lester Solbakken

​​​19:00 Coding agents demonstrate the future of agents?

​Coding agents show how retrieval sits at the core of effective agents that actually gets work done. This talk first explores how retrieval is central for coding agents, how coding agents use retrieval in their loop, and then discusses how these patterns can generalize to other domains. 

​​​Presented by ​Jo Kristian Bergum ​

​​19:30 A panel with Oscar, Jo and Lester

​What about long context models? Is RAG dead? Agents everywhere. What is the difference between a workflow and an agent? And does it matter? ​​We’ll open the mic.

​​​20:15 Drinks & debrief 

​​​Corner the speakers or network with other builders.