On September 15, I had the opportunity to attend the AWS Community Day DACH in Berlin. The event provided numerous insights into current developments in cloud computing, artificial intelligence, and modern software development. It was particularly noticeable that many presentations focused on the future topic of autonomous AI agents. The main topics included Context Engineering, Loop Engineering, Agentic Coding, as well as the secure execution of AI-generated code in sandboxes and MicroVMs.
The first presentation, “Context Engineering: Why Your Agent’s Memory Is an Architecture Decision,” focused on the question of how AI agents can access the right information. Using a practical example, it was demonstrated that despite having correct documentation, an agent was unable to properly assess a production incident because it lacked information about the current state of the system. The key message was that modern AI agents cannot rely exclusively on knowledge bases or vector searches. Instead, they need different data sources for different types of context, such as current operational data, historical information, or documentation. This made it clear that Context Engineering will play an important role in the future development of powerful AI systems.
Afterwards, I attended the presentation “AI Agents Need Sandboxes: Why AWS Built Lambda MicroVMs.” This presentation addressed the challenge of allowing AI agents to safely work with and execute code autonomously. As agents increasingly take on tasks that go beyond simply generating code, they need to be able to start programs, process data, and perform calculations. AWS introduced so-called Lambda MicroVMs for this purpose. These isolated runtime environments make it possible to safely execute code generated by AI without putting other systems at risk. Particularly interesting was the live demonstration, in which an agent independently executed code inside an isolated MicroVM. The presentation made it clear that sandboxes and MicroVMs will play a central role in the future as AI agents take on more and more responsibility.
I concluded my day with the presentation “Every Builder Ships: Scaling Agentic Engineering with Kiro at the Bundesliga.” This presentation focused on how AI agents can be used in professional development teams. While many demonstrations show how individual developers can work faster with AI, this presentation looked at scaling these capabilities across entire teams. The Kiro platform was presented as a way of controlling agents through rules, specifications, skills, and automated validation mechanisms.
The concept of Loop Engineering was particularly interesting. Unlike traditional Prompt Engineering, where the quality of the result strongly depends on a single input, agents here work in continuous feedback loops. They analyze a task, create a solution, run tests, evaluate the results, and then independently optimize their work. This process is repeated until the desired quality requirements are met. As a result, the role of developers is increasingly shifting from direct programming towards defining processes, rules, and quality standards.
A clearly recognizable trend across all of the presentations I attended was the shift from traditional AI assistants towards Agentic Coding. In the future, AI systems are not only expected to provide suggestions but also to independently work on tasks and prepare decisions. However, this requires three essential components: the right context through Context Engineering, continuous improvement mechanisms through Loop Engineering, and secure execution environments in the form of sandboxes or MicroVMs.
Overall, the AWS Community Day DACH was a very interesting and informative event for me. In particular, the presentations on Agentic Coding, Loop Engineering, and secure runtime environments demonstrated how strongly software development is currently changing. The concepts presented showed that AI agents will play an increasingly important role in the future and that developers will increasingly be responsible for creating the right framework and conditions for their use. The Community Day therefore provided an exciting outlook on the future way of working in modern software development.
