Homborsund AI Summit

Michael recently became a father. His child will never remember a world without AI in it — will never not know the feeling of talking to something that answers back. What does that childhood look like? What do we teach, what do we protect, and what becomes of curiosity when every question has an instant, confident answer? A quieter talk from someone who has suddenly started thinking about the long term.

Everyone was excited at the kickoff. The pilot worked. The demo got applause from people who don't usually applaud. And then — nothing. The budget moved, the sponsor changed roles, and the Slack channel went quiet. Robin investigates the scene of the crime, talks to the usual suspects, and asks why organisations keep murdering projects that were technically alive.

Everyone has a prompt that works a little too well and one that fails for reasons no one can explain. Bring yours. We trade tricks, superstitions and half-working incantations — and then ask the harder question: when the models get better at reading us than we are at writing to them, what is a prompt, really? And what comes after?

If AI can write, diagnose, plan and execute — what exactly should a person spend their Tuesday doing? Lars takes on the largest question of the decade with the calm demeanor of someone who has already made peace with several possible answers.

Short answer: we probably don't. Long answer: when models can reason, remember and spin up exactly what you need based on who you are, the whole concept of an 'agent' starts to feel quaint. Øyvind makes the case for retiring a buzzword before it even peaks.

The real talk happens after the talks. Dinner, demos, bonfire debates and the kind of half-formed ideas that only surface when the projector is off and the sea air hits. This is the part no AI can replicate. Yet.