Software Architektur im Stream

Moderne LLM's generieren oder überarbeiten ganze Applikationen in Rekordzeit. Die Frage ist, wie sich Architektur und Struktur auf die Qualität und die Wartbarkeit der Ergebnisse auswirkt. Spielt eine saubere Struktur noch eine Rolle? Was passiert, wenn die KI ein Problem nicht beheben kann?

We are rapidly moving from AI as a tool to AI as a participant.

Agentic systems now write code, review designs, raise flags, and hold context across long-running tasks. We are, whether we intend it or not, building teams that include non-human members.

But what is a team, really? Not a group. Not a workforce. A team is something older and stranger. Perhaps it begins with family, with the commitment to show up even when it is costly, with the shared meaning that makes humans build, play, march, dance, and sacrifice together.

In this episode, Avraham Poupko and Ralf D. Müller explore what it actually means for an AI to be a teammate rather than a tool.

Wir stellen einen Menschen und eine KI vor dieselbe Architekturfrage und bitten beide, ihre Empfehlung kurz und knackig zu formulieren: mit Architekturentscheidung, Begründung, Diagramm und bei Bedarf einem kurzen Code-Beispiel.

Diese beiden Entscheidungen legen wir euch vor, natürlich ohne zu verraten, wer welche geschrieben hat. Dazu jeweils zwei Fragen:

Welche Empfehlung stammt vom Menschen?

Welche Empfehlung würdet ihr in eurem eigenen Projekt umsetzen?

Damit die ADRs greifbar bleiben, arbeiten wir an einem durchgehenden Beispielprojekt, das mit jedem ADR weiter wächst.

Am Ende interessiert uns vor allem eins: Wie gut trifft die KI eine Architekturentscheidung, wenn sie dieselben Fakten vor sich hat wie wir?

Dies ist die Keynote für die Software Architecture Info Days https://www.infodays.de/sa/programm/konferenzprogramm/details/keydi-1

In this special episode of Software-Architektur im Stream and maschinenraum.fm, host Lucas Dohmen sits down with web development veteran Dylan Beattie to explore the surprising power of modern CSS. Long dismissed as a mere styling tool, CSS has evolved into a robust language for layout, animation, accessibility, and complex interactions — often eliminating the need for heavy JavaScript frameworks. Drawing on decades of experience building for the open web, Dylan shares how combining semantic HTML with contemporary CSS can simplify front-end architecture, reduce dependency on frameworks like React, and help developers ship faster and more maintainably. Tune in to discover how web development is being reimagined.

Wie betreibt man eine Webanwendung in der Simple Cloud – also auf transparenter Basis-Infrastruktur (VPS, Networking, Firewalls) ohne Cloud-Managed-Services? Mit Dirk Breuer bespreche ich, wie unser Setup bei fejo.dk, einem der führenden Portale, um Ferienhäuser in Dänemark zu buchen, funktioniert und wie wir dabei auf Einfachheit gesetzt haben. Wir gehen das Setup durch: von Terraform über Ansible zu Docker und Capistrano. Wir freuen uns auf eure Fragen vorab auf Mastodon oder LinkedIn oder live bei der Aufnahme.

In this episode Sander Hoogendoorn talks with Eberhard about his experience at iBOOD with AI-supported software development. Why abandon Pull Requests? What does AI-assisted coding review really change? And what happens to collaboration when AI is used?

We also look beyond the tools: How does AI change the role of developers, software architecture, and technical leadership? If developers increasingly own outcomes rather than keystrokes, what skills and practices become important?

And perhaps most importantly: What does it take not just to survive the AI-native era, but to thrive in it?

Sander will also talk about this subject at Software Architecture Gathering (2026-11-16 - 19, Berlin). Use code SATV_15 for 15% off.

Kelsey Hightower is one of the most influential voices in the Kubernetes community. During his career, he worked in a variety of engineering roles at Puppet Labs and CoreOS before joining Google, where he became a developer advocate and engineer working with Kubernetes.

In this episode, Kelsey talks with Eberhard about the evolution of modern infrastructure, Kubernetes, and what these changes mean for software development more broadly. They explore the ideas, practices, and lessons that matter when building and operating software in an increasingly complex world.

Links

• GopherCon 2017: Self Deploying Kubernetes Applications - Kelsey Hightower

Am TechRiders Summit 2026 (https://tech-riders.de) hatte Softfware-Architektur im Stream die Gelegenheit, kurze Gespräche mit einigen Speakern zu führen. In diesem Special spricht Petra Clemens (CIO Eppendorf) mit Hansjörg Gude (Geschäftsführer SWAGLab) über aktuelle Herausforderungen in der IT Strategie und auch Diversität in der IT.

Mehr zu Petra Clemens:

https://de.linkedin.com/in/petra-clemenshh

Psychological safety is essential for teams to speak up, disagree, make mistakes, and learn. But it is not about making everyone comfortable — it is about feeling safe enough to be uncomfortable.

In this episode, Gitte Klitgaard explores with Eberhard the role of leadership in creating psychological safety, how to recognize when it is missing, and what can be done to improve it.

Gitte will also talk about this subject at Software Architecture Gathering (2026-11-16 - 19, Berlin). Use code SATV_15 for 15% off.

For 30 years Randy Shoup has helped shape engineering at some of the industry’s most demanding companies - eBay, Google, Stitch Fix and WeWork among them - building highly scalable, distributed systems that serve hundreds of millions of users; today he leads engineering at CircleCI. He is also a long-time speaker and writer on how healthy engineering organizations really work: small multidisciplinary teams, autonomy and accountability, goals over solutions. He joins Ralf D. Müller for an hour on the question our whole field is circling right now: what does generative AI really change about how we build software, and what does it leave exactly where it was? One of Randy’s long-standing lines is to “model the world as it really is, not the way you would like it to be”; we carry that into the AI moment, asking where teams keep building the world they wish they had, and where Randy nods along with the current AI hype versus where he pushes back.