mkdev — DevOps and Cloud Native Consulting

What the Hell Happened to IT in the Last 100#? | ✉️ #100

Illustration for mkdev dispatch #100 showing a bearded man with glasses beside the text “What the hell happened to IT in the last 100#?” against a paper-plane patterned background.
Updated: | Published: 26/09/26

Hey! 👋

We made it to 100 dispatches. I thought about doing a proper anniversary recap, but looking back at the last two years in IT, I mostly have one question: what the hell just happened?

Think about the jump from asking a chatbot for a script to giving an agent access to your repository and letting it work through a list of tasks. Kirill recently described how his working setup barely changed for a decade, and then AI had him switching tools constantly. That sounds about right. We’re changing how we work while still trying to finish the work we already have.

And it wasn’t just code. DeepSeek released reasoning models you could download yourself. ChatGPT gained native image generation. Veo 3 brought video with generated dialogue. Suddenly, a conversation about automating some programming tasks became a conversation about almost everything involved in making a digital product. I’m excited about that, but I don’t think we’ve really had time to absorb it. There’s always another release to try.

Of course, Kirill also reminded our readers that I’ve been predicting the disappearance of developers and DevOps engineers for years, while still working far too much myself. Fair enough, Kirill. But I’m not convinced that “we’re still busy” settles the argument. If a small team can do what previously required a much bigger one, that changes the business, even if nobody changes their job title. I wouldn’t promise someone starting their career that everything will be fine just because we still have tickets in the backlog.

Meanwhile, the infrastructure has an excellent way of keeping us humble. In one of our projects, an AI-generated implementation produced over a million Lambda executions through an unnecessarily complicated polling setup. The task was supposed to reduce costs. I quite like that example because it captures the awkward place we’re in: these tools are useful enough to trust with real work, and that trust can get expensive when we stop paying attention.

So I’m not going to use #100 to predict exactly what our industry will look like in another two years. I’d probably regret it by #110. There’s plenty worth building now, and plenty we still need to understand. Thanks for reading, replying, and following along while we try to figure it out ourselves. And tell us: looking back at your own work, what do you do differently now that would have sounded ridiculous two years ago?

What We've Shared

  • ​How GPT-6 Astra Killed My Ruby on Rails App: Kirill handed GPT-6 Astra a long-postponed migration of mkdev.me from Ruby on Rails to Hugo and let it run with surprisingly little supervision. The result worked — but the path there was full of overengineering, impressive autonomy, missed details, and a few good lessons about what today’s most capable coding agents are actually like to work with.

  • ​Grok Bot is a better product than OpenClaw and Hermes Agent: Kirill also takes a closer look at Grok Bot and why it feels so pleasant to use even when its underlying AI isn’t necessarily better than the competition. He breaks down the product decisions that make the difference, from simple Bots and built-in computers to multi-account plugins and smoother browser interactions, while also pointing out where Grok Bot still has room to improve.

What We've Discovered

The 101st mkdev dispatch will arrive on Friday, October 9th. See you next time!