Hello, this Sergei.
Last week, I watched a documentary about the early days of punk music.
It all started with a DIY approach. No good instruments to play? DIY and play on whatever you have. No studio to record your music? DIY — press the “Record” button on your tape recorder. No money to buy “cool” clothes? Screw it, do it yourself. And this simple idea completely reshaped the music industry and influenced almost every kind of music that came afterwards.
While watching the documentary, I was thinking about AI as it is today. I’m not sure the AI industry is developing in the right direction, but for now, I believe the most valuable way to work with AI is not to generate shit, but to use it to help us do cool things. And here is the punk approach fits best. So for now the way how some of us use AI is a sort of reflection of that punk approach: stop buying expensive things (they’re expensive partly because of the privileged position of the companies selling them — they have the facilities to produce them and the money to market them) and just make them yourself.
From time to time, I see diagrams showing the amount of software developed over the last 3–4 years alongside the amount of money and users those products have. The growth in the number of apps doesn’t correlate with the number of people using them. Some analysts use this as an example of “vibe coding producing a lot of apps nobody uses.”
I think that’s the wrong angle. It absolutely doesn’t matter how many people use those new apps. For me, it means two things.
1. The software market was ill and had been captured by a few big players. It underrepresented the actual needs of users. So with AI, people have started making the apps they really need — even if there is only one user who installs it: themselves. This is my approach actually.
2. Maybe we’re entering the end of the cloud-based-everything era. Local again. DIY. Build it, use it, own it. Keep your data with you.
So we can still be punks. Like Malcolm McLaren and Vivienne Westwood in the mid-’70s.
New on AI For Newsroom this week
Stories, guides, initiatives, and signals we surfaced in this issue.
Dystopia, Inc: Would You Like Your News With or Without Hallucinations?
One in five Brits using AI to keep up with news
Big If True
Agent Office
How I built an agentic system that evaluates 200 newsletters a day
What 99 AI licensing deals reveal about the future of publishing
What AI actually looks at when judging a news publisher
Arc XP launches Compass to personalize news home pages
Axios says AI is breaking the link economy
We Need Google Trends for AI
The Economist Group’s AI Hub
Comscore data shows how AI discovery is splintering beyond ChatGPT
AI slop will not save legacy newspapers
Is republishing content worth it? A look at SEO, user needs and syndication amid AI disruption
Value in the emerging AI-mediated information ecosystem
Automated podcast transcription workflow
Custom “vibe-coding” GPT
Get the weekly email digest
One curated Friday roundup for newsrooms. No spam — you can unsubscribe anytime.