What AI is changing in the talks we receive
The 2026 call for papers has changed face. What the proposals we received say about the state of the profession.
We went through the proposals received for the 2026 edition. The result says something interesting about the state of the profession.
Fewer demos, more production
Two years ago, AI topics revolved around “look what we can do”. This year, most proposals talk about what breaks: runaway costs, unworkable latency, answers impossible to evaluate, guardrails bypassed.
It's the sign of a technology leaving the demo phase. People have put things into production and are coming back with scars.
AI spills over beyond AI sessions
Another finding: a good share of talks labelled architecture, DevOps or security now include an AI component. Not as the main topic, but as a component of the system being described.
That's precisely why we don't have an “AI” track that would swallow everything. The subject has become cross-cutting.
What hasn't changed
The level of rigour expected on stage. A proposal that explains what an LLM is gets rejected, however good the speaker. The audience already knows.
Seven rooms, forty talks, one single day
Building the DevDay grid is mostly about trade-offs. How we go about it, and what we systematically turn down.
Forty talks, seven rooms, nine hours. On paper it's a spreadsheet. In reality, it's three months of trade-offs.
We start with the rejections
Every year, the call for papers brings in far more proposals than slots. So the grid isn't built by picking the best: it's built by setting aside everything that looks alike.
Two excellent talks on the same topic is one talk too many. We keep the one that brings lived experience rather than an overview.
Balancing the rooms
Seven parallel rooms means seven simultaneous choices at every slot. If two highly anticipated sessions land at the same time, we've failed at our job. So we spend as much time on scheduling as on selection.
The internal rule is simple: at no point in the day should an attendee have zero relevant sessions for their track.
What we don't do
No sponsored session disguised as a technical talk. No product keynote. A sponsor can take the stage, but with a topic that passes the same filter as all the others.
It's the kind of rule that costs money and keeps people coming back.
Why DevDay has three tracks instead of a single programme
Developers, AI Builders, Executives: the split isn't a marketing gimmick. It comes from what we've observed edition after edition.
The question comes up every edition: why not run a single programme and let everyone pick? We tried. It doesn't work, and here's why.
The problem with a single programme
When everyone looks at the same grid, two things happen. Developers miss in-depth technical sessions buried among strategic talks. And executives, lacking landmarks, end up in a room where code scrolls by for forty minutes.
Nobody wins. The speaker loses half their audience, and the audience loses an hour.
Three professions, three sets of needs
A developer comes for technical depth and production war stories. An AI builder faces questions of evaluation, cost and guardrails the former hasn't encountered yet. An executive makes decisions on budgets and hiring without needing to see a line of code.
These aren't different levels of the same topic. They're three professions with three different lenses.
What the split doesn't do
It doesn't lock you in. The ticket gives access to the whole event, and the rooms are open. The track is a compass, not a barrier: it tells you where you're most likely to find what you're looking for.
In practice, many attendees move between two tracks during the day. That's exactly how it's meant to be used.