WTF, AI? · Leuven

01

Why

Something changed this year in who is asking. People who have never had an opinion about machine learning now want to know whether the thing they read about was real, how worried to be, and whether anyone competent is handling it. Those are good questions. There is nowhere in Leuven to ask them out loud.

The reflex, hearing something like this for the first time, is to file it away on the assumption that someone else must surely be on top of it. Most of the time that is the right call, because you cannot take everything seriously. But it is worth knowing when it is the wrong call, and that is not something you can work out on your own.

Reading more does not fix it either, and I say that as someone who tried for years. You can follow the arguments closely and still have no idea whether your own reasoning holds, because nothing has ever tested it. That only happens in front of people who know the material and are willing to tell you where you are wrong. Almost everyone I know who ended up doing serious work on this got started in a room, not in a feed.

So: one evening, a room, about ten people who know different parts of this well, and whoever turns up. If it works I will do it again.

The format comes from AGI.WTF in Berkeley, which was announced five days ahead and filled. This is smaller and unaffiliated.

02

The year so far

If the volume of AI news this year outran your sense of what was real, start here. Everything on this line has a date and something behind it. The sources are named; the links go up as each one is checked against the primary document rather than the coverage of it.

Scroll horizontally for the full timeline

Read the whole dossier →

03

What happens

Half an hour from me on what actually happened this year, dated and sourced, so the room starts from the same facts. After that, whoever among the guests feels like answering sits at the front and takes whatever you ask. They come from different directions, and they swap out whenever they want to. The rest of the time they are in the room with everyone else.

Nobody is assigned a slot and nobody is stuck in a chair, guests included. Ask your question at the front, then go find the person whose answer you did not buy.

18:30
Doors
Come early if you can. The people standing around before it starts are half of what you came for.
19:00
What actually happened in 2026
Thirty minutes, from me. No predictions, just dated events with sources, and what can and cannot be read into them. The Hugging Face intrusion in July. The Navier–Stokes proof in September. What Washington and Brussels have each done since.
19:30
Questions start
Whoever wants to answer sits down at the front and takes them. No talks, no slides, no agreed line. If the people up there disagree with each other, you get to watch that happen, which is more useful than any one of them being right.
20:30
They swap, you scatter
People get up, others sit down, whenever they feel like it. The ones not up front are in the room in small groups. Nobody is running the changeover, and you can move around just as freely.
22:00
Out
Whoever is still arguing moves to a bar nearby.

04

Who is up front

To be announced. I am inviting a small number of people who each know a different part of this well: research, policy, journalism, people who build the systems and people who study what they do to us. Names appear here as they confirm.

to be announced

05

What this is not

You will not leave with an answer. Some of what you hear will be confidently wrong, and on the night you will not be able to tell which parts. That includes what I say. I have been following this closely for years and I am still unsure what is actually worth doing, so push back when you think I am off.

What an evening can do is give you enough checkable facts to start building a view, and enough sense of who is working on what to judge whether someone competent really is on it. If you decide they are, go home and think about something else. That is a real outcome and a fine one.

06

Who it is for

Anyone with a question. Researchers, students, people who build software, people who do not, your parents, the colleague who keeps raising it at lunch. The basic questions are welcome and usually the ones the rest of the room was quietly waiting for.

If you already feel oriented, come anyway and spend the evening helping other people get there. Free, and open to whoever turns up.

07

Practical

When: The date is being set around the people joining the panel. Sign up and you get it first. It will be 19:00 to 22:00, doors at 18:30.
Where: The room is being confirmed. Sign up and you get the address as soon as it is fixed.
Language: English
Cost: Free
Up front: A few guests, whoever feels like it, swapping when they want
Host: Mara Masaeva

08

Come

Sign up so I know how many chairs to find.

Why

The year so far

What happens

Doors

What actually happened in 2026

Questions start

They swap, you scatter

Out

Who is up front

What this is not

Who it is for

Practical

Come