Updated October 2026: Dílna.AI launched as an open, free community without paid tiers. The article has been updated to match.
The Tab Problem
Last Tuesday, I had 34 browser tabs open. All AI-related. A new ChatGPT model announcement. Some guy on X is claiming Claude just replaced half his team. A paper about code generation benchmarks that contradicted everything I’d read the week before. Three YouTube videos saved in “watch later” — where videos go to die. An n8n workflow template that looked promising but required an API I’d never heard of.
I closed all of them. Went back to work. Shipped the feature I’d been building.
Then Wednesday came, and I did the exact same thing.
This has been my life for the past year and a half. I suspect it’s been yours too. The amount of AI-related news, tools, techniques, and hot tips that show up every week is genuinely impossible for one person to keep track of. And I don’t mean “difficult.” I mean, impossible. There are people whose full-time job is writing AI newsletters, and even they miss things.
But here’s what actually bothers me. It’s not the volume. It’s the uncertainty. When I see a new tool or technique, I can’t tell whether it’ll save me 10 hours a week or waste an entire weekend. The signal-to-noise ratio in AI content right now is maybe 1:20 on a good day. And most of the “signal” comes packaged in breathless Twitter threads that make everything sound like it’ll change the world.
I needed a filter. Not an algorithm. A human filter.
What Happened When I Started Talking to Other Devs
About six months ago, Karolína Vyskočilová and I started having regular calls. Karolína runs her own WordPress and WooCommerce business — she’s been in the game for 15+ years, originally came from linguistics, and has been using Claude Code and other AI tools daily since they became usable.
The calls started as informal check-ins. “Hey, have you tried this?” “Yeah, it’s hype.” “What about this workflow?” “That one’s actually great — here’s how I modified it for client work.”
Within a few weeks, we’d saved each other probably 20 hours of trial and error. Not because either of us is smarter than the other. Because two people covering different ground and comparing notes is exponentially more efficient than one person trying to cover everything.
That’s when it all clicked. If two people sharing notes saves this much time, what happens when ten people do? Twenty? A hundred?
That’s where Dílna.AI is heading. Not from a business plan or a launch strategy. Given that this works, more people should have access to it.
Why a Community and Not a Course
We actually had a course planned. Dílna.AI was originally going to be a structured online course about using AI in professional development workflows. We’d built the curriculum, outlined the modules, and figured out pricing.
Then we realised something that should have been obvious from the start: a course about AI tools is outdated before you finish recording it.
We recorded a module about prompt engineering patterns in December. By January, two of the five patterns were no longer relevant because the models had become good enough that they no longer needed them. By the time we’d edited the video, Claude had shipped three major updates that changed how half the features worked.
A course is a snapshot. AI development right now isn’t a snapshot — it’s a live feed. What we actually needed was a format that could keep up.
A community keeps up. Every call can reflect what happened that week. Every Slack conversation can be about what someone tried today. The knowledge stays current because the people in the conversation have current information.
And truthfully? The most useful thing in our calls with Karolína was never a prepared presentation. It was one of us saying, “I tried X on a real project and here’s what actually happened.” That kind of feedback doesn’t come from a course. It comes from a room full of people who work with AI.
What We’re Building
A Czech-language community for developers who use AI in their daily work. That’s the one-line version.
The longer version: we run live calls every two weeks where we break down the most important AI news and tools. Not everything — the stuff that matters for people who write code for a living. I’m already building an automated pipeline that pulls from about 50 sources, scores everything for relevance, and gives me a pre-filtered list to review. The actual human preparation will take about an hour per call instead of the full day it takes now when I try to keep up on my own.
Each call follows a structure. Top headlines with context — what happened and why you should care (or shouldn’t). One deep examination of a specific tool or technique with a live demo. A section where members share their own discoveries. Open discussion. And at the end, a list of things to try before next time.
Between calls, there’s a Slack workspace for sharing tips, asking questions, and — knowing developers — arguing about whether AI-generated tests are actually useful. (They are, with caveats. It has already become a recurring debate.)
We don’t send a newsletter. There’s already plenty of generic AI content, and we don’t want to add to it. After each call we publish a short summary on dilna.ai for anyone who couldn’t make it.
The whole thing runs in Czech because that’s our audience. Czech developers solving Czech-market problems, discussing things in a language that lets them be precise and natural, rather than showing off their English fluency.
Open, Free, No Tiers
When we started, we planned paid tiers on top of the free community: a Professional membership and a small Mastermind group. Once the calls were actually running, we dropped that idea.
The reason is simple. The most valuable thing in Dílna.AI is the people who show up and share what they tried. A paywall would split that group in two, and the conversations would get poorer on both sides. Membership communities also have a trust problem: too many of them sell access to a Slack channel that posts two messages a week and call it "premium". We didn’t want to be one of them.
So Dílna.AI is open to everyone and free:
- Public Slack channels where you can ask, share and discuss whenever you need to.
- Live calls every two weeks where we talk about what we actually use, swap experience and give each other tips.
- Meeting in person for a beer and dinner now and then, to cover what we didn’t get to on calls and Slack, and to get to know each other.
There’s no corporation or marketing agency behind it, just two independent developers with real commercial experience. No credit card, no trial that quietly turns into a subscription, no upsell.
Who This Is For (and Who It Isn’t)
The ideal Dílna.AI member is a developer or tech lead who’s already using AI in some capacity and wants to get better at it — faster. Someone who’s tried Claude, Codex, or Cursor but suspects they’re barely scratching the surface. Someone who reads AI news but can’t tell what’s worth their time.
It’s specifically for people who build things. React, WordPress, Laravel, Python — we don’t care about your stack. We care that you’re shipping real projects and want AI to help you do it better.
It’s probably not for you if you’re looking for a beginner’s guide to “what is ChatGPT.” We assume baseline competence. The conversations will move quickly, and nobody’s going to explain what a prompt is.
It’s also not for you if you want someone to tell you AI is going to replace all developers. We don’t believe that, and that debate isn’t interesting. AI is a tool. A remarkably powerful one that’s changing how we work, but still a tool. Dílna.AI is about getting better at using it, not philosophising about whether it’ll take your job.
What Our Calls Already Taught Me
Even just the regular calls with Karolína have changed how I think about AI in my own work.
Before we started talking, I evaluated tools in isolation. I’d try something, form an opinion, and move on. Now I hear how someone else used the same tool on a completely different type of project, and my understanding is dramatically more subtle. Cursor is incredible for greenfield React work, but it struggles with legacy PHP codebases. Claude Code handles large codebases better than anything else, but needs very specific instructions for WordPress conventions. GitHub Copilot is still the fastest for autocomplete, but falls behind on architectural decisions. Codex is strong at UI design.
None of those is opinions I could have formed alone. They came from comparing notes with someone doing different work.
Scale that up to a room of twenty, thirty developers — each working on different stacks, different project types, different client demands — and the compounding effect is what I’m betting the whole community on.
Join Us
The community is up and running. We’ve held four calls since July 2026, and you can read the summaries on dilna.ai. Joining is free, with no obligations.
Head to dilna.ai and fill out the short form. You’ll get access to the public Slack channels and an invitation to the next call.
I’ll be there. Karolína will be there. And if our calls so far are any indication, at least one person on every call will say something that saves you a weekend of pointless experimentation.
That’s worth showing up for.
