Erwin Blom: “AI only gets interesting once it’s in my hands”

Erwin Blom is a media maker and internet pioneer. He was there when video on the web was still a postage stamp that looked like nothing at all, built podcasts and formats at VPRO among others, and these days more or less lives in his AI tools. We talked to him about punk, house and the early internet, and about why AI has come along at exactly the right moment for him.

You were there at the start of the internet, but also of punk, hip-hop and house. Are you glad you had to figure it all out without AI, or would you rather have started now?

There are two sides to that. I think it’s valuable that I know where we come from. That I was there when video was still a postage stamp that looked terrible and audio didn’t sound like anything yet. It makes me appreciate what’s happening now all the more. At the same time, that history can also be dead weight. If you’re sixteen or twenty now and you start without that baggage, nothing is sacred to you. But personally, I love that I lived through all of it. The beginning of something, whether it’s music or technology, is almost always the most fun.

“Something only becomes real to me once it’s in my hands.”

What makes a beginning like that so much fun?

The market hasn’t settled yet, there’s still all sorts of stuff left to invent. In the early days of house, almost every record was good, because the people making it were in it out of fascination, not for the money. Then come the hangers-on who want a piece of the action, with weak knock-offs. You’re seeing the same thing with AI now. Almost all the big names in AI are on what used to be Twitter; they say things there, they feel close. I had that feeling of closeness at the start of the internet, at the start of Twitter, and now again with AI. It fades on its own, but then something new comes along.

You come from the punk scene yourself, which is known for its do-it-yourself mentality. A lot of people see AI as the exact opposite: you leave the work to the machine. Yet you embrace it. Isn’t that remarkable?

If you’re creative, that’s exactly when you should go looking for what you can do differently with the technology. I don’t even think it has anything to do with age, by the way; musicians of twenty-five are often against it too. I find that resistance hard to understand anyway. Look at house: that was just a drum machine that was supposed to sound like a drum kit but couldn’t, so it sounded like something new. Or a 303 that was meant for bass lines, but people completely abused it, and suddenly you had acid. That’s how you should see this technology too. On the one hand it makes things easier, on the other you have to look for how it helps your creativity.

So how do you use it yourself?

I practically live in Codex. Whether you pick Claude or Codex doesn’t matter that much, they do similar things. All those companies are trying to become the most important app on your computer; there’s even a browser built in. For someone like me, not a programmer but a knowledge worker, this is fantastic. I dig into subjects and pass that knowledge on, through a newsletter, a podcast, a talk or a training session. You’re basically building your own operating system, with all your knowledge in it. I’m writing a book right now. If I come across an article, I put it in and ask: does this add anything to what I already have, or is it already covered better somewhere? Often it says no, and I throw it out. That way there’s always someone working along with me.

A lot of people say: so you’re having someone else do it.

Partly. At VPRO I had a team of twenty-five. I never wrote a line of code there myself and never did any design, I was the boss. Still, it felt like I was helping make it. This feels exactly the same. I’m the one who keeps saying: no, do this differently, this is how I want it. Taste and vision are what make the difference.

You’ve worked a lot with designers and programmers. How do you feel about no longer hiring them?

People say the technology is putting people out of work. To some extent that’ll be true, but I think it also creates a lot of work. A programmer gets to eighty percent quickly, but that last twenty percent is complicated and takes a lot of time. That’s what you have to specialize in, security for example. And designers can suddenly afford a programmer they’d never have had otherwise, and vice versa.

At VPRO we had programmers and project managers, and every idea quickly started at fifty or seventy thousand euros. Often I didn’t have that money. Those same ideas I can now actually carry out. Soon there’ll be loads of small companies springing up alongside us. You make a first version, see whether there’s an audience for it, and then it’s worth putting money into. But you still need people who really know a field and understand what people are missing.

“What can I do now that I couldn’t do before? That’s a far more interesting question than faster or cheaper.”

You said earlier that ‘faster and cheaper’ actually bores you.

Efficient, sure, I suppose. Nice when it happens, but boring. A much more interesting question is: what can I do now that I couldn’t do before? Recently I came across a VPRO series from 1994, Het Web, about the early internet. And I thought: what could I do with that using today’s technology? Within a day I had a site up, with articles that compared then and now based on the transcripts. I cloned my own voice with ElevenLabs, so anyone who’d rather listen than read can hear me reading it out. With NotebookLM I made videos to go with it, and I had an epub generated automatically for anyone who wants to read it as a book. One day. VPRO itself puts seven videos online; my site has those same seven plus all that extra value. I think that’s insane.

You also collect stories like this, from people who build things themselves.

Yes, all of them people who aren’t programmers or designers, but who make a solution for their own problem. Someone built great tooling for hockey teams. When it suddenly started snowing in February, someone made a crowdsourced snow map, because you can’t plan snow: when it’s there, you have to get out on it right away. Nobody would put money into that, because you don’t know if it’ll happen again next year. Someone else maps out all the supermarket discounts. You name it. And this is only the beginning.

Would you venture a guess about where we’ll be in two years?

The smartest people at the big companies don’t even dare say anything about three months from now. But you can see how fast it’s getting better. At first you had to prompt very precisely; now you describe the end goal and the tool works out the route to get there. Security still plays a role, but tools like Lovable see it as in their own interest to send you out into the world as safely as possible. I think we’ll soon have a kind of all-encompassing operating system through Claude, Codex or Gemini, with agents that go to work for you at set times or on certain signals, combined with whatever you think is important at that moment. Sometimes in the middle of my work I’ll quickly ask for a dashboard, just to see how far along I am. Not for the outside world, purely for myself.

Don’t you have any concerns?

One. Maybe it’s going too fast. When electricity was developed, it took a long time before people had it at home. AI is moving insanely fast, and I wonder whether people can take it in. Much more is already possible than what people are talking about. It reminds me of the MiniDisc: a better format than the CD, because it didn’t skip and you could just delete tracks you didn’t like. But after vinyl and the CD, you can’t ram the MiniDisc down people’s throats as well. Sometimes it’s just too much. The technology is great, but how much people can handle remains to be seen. Someone said recently: if you’re outside the bubble and you do something with AI for an hour a day, you’re already ahead of ninety-seven percent of people. That’s the reality.

Finally, the thoughtchain question: who should we talk to next?

Bart Brouwers. He used to work at De Telegraaf and at Dichtbij, and now runs a company together with Marian van Houten. They previously founded punt.nl and do a lot with AI: they had an AI editor early on and make automated news sites around themes and around places that no longer have their own newspaper. Bart is also a professor of journalism in Groningen, so he’s been along for the whole ride, from the nineties until now.