Insights - Many talk about AI Engineering – few show how.

Sunday at the Agentic Conf 2026 in Hamburg – and a feeling I hadn’t had in a long time. No vendor stage, no business fluff. Just people talking about things that genuinely matter to them. Like the “old” unconferences (Judith Andresen). That kind of energy has become rare. We are all pioneers. Nobody has the handbook – we are writing it together, right now. A few insights that ran through the whole day:

The models are good enough. The bottleneck is no longer code generation – it’s decisions, requirements, acceptance criteria, whether TDD or SDD. An “AI Harness” is non-negotiable in AI Engineering. Many talk about it; Björn Rochel showed in his talk what it can actually look like.

Much of what we have learned over decades still holds. Focus. Clear requirements. Except that focus today means context. And #SDLC remains important – just in much shorter cycles, daily, multiple times a day.

AI as a personal sparring partner. Kaizen, every day. Tereza Iofciu, PhD showed in her Workshop how that works – and her Nemesis Starter Kit makes exactly that possible.

How do you scale with AI in development? Benedikt Stemmildt 👨🏼‍💻🧙🏼‍♂️ and Leo Peters showed a concrete approach – the Factory – live using the Tegtmeier case study.

Tools like Terraform are starting to feel increasingly sluggish. Paul Stack nailed it. SWAMP – is on my list to try.

Don’t wait. Start, learn, optimise. AI doesn’t just help with coding.

Thank you to the organisers Stefan Munz, Sebastian Korfmann and Sebastian Reimann for creating this space! 🙌