Articles and insights on AI consulting, software engineering and security, from our founder in Vienna
By AIPathTech Editorial · 16 November 2026 · 3 min read
A rule was written down, read that morning, and broken that evening. No error was raised, because no error occurred. This is the failure mode nobody sells you a tool for.
By AIPathTech Editorial · 9 November 2026 · 4 min read
A rule was applied to a hundred blocks, correctly, in a fraction of a second. In one of them the phrase it removed was not a label but the subject of the paragraph. Nothing failed, nothing went red, and the text still read as English.
By AIPathTech Editorial · 6 November 2026 · 4 min read
A function was placed inside another function. Everything parsed, everything looked correct, and the page would have failed for every visitor. It never reached anybody, because the command that publishes the site builds it first and refuses to send a broken build.
By AIPathTech Editorial · 5 November 2026 · 3 min read
Most companies that say a human stays in the loop mean somebody reads the output before it goes out. That is not a human layer. That is a proofreader.
By AIPathTech Editorial · 2 November 2026 · 5 min read
AI writes code here every day, and it is wrong often enough that a company which trusted it would already have broken something. Both sentences are true, and everything we do about governance exists because both are true.
By Maria Catalina Kovacs · 28 October 2026 · 4 min read
We built a page that shows how many people read what we write. It was correct, it worked, and for weeks it was opened about as often as a filing cabinet. Nothing was broken. That is what makes it worth writing about.
By AIPathTech Editorial · 26 October 2026 · 5 min read
We wanted one number: how many people opened a page. The server already knew it. Reading it back turned out to be a paid plan, and the price was not the reason we wrote our own.
By Maria Catalina Kovacs · 23 October 2026 · 3 min read
Our site rebuilds every morning, so the easy implementation would announce that all thirty six of our pages had just changed. Thirty six lies a day, told with perfect sincerity by a shell script.
By Maria Catalina Kovacs · 21 October 2026 · 7 min read
We needed numbers and we did not want to track anybody. Those two sentences are supposed to be in conflict. They are not.
By Maria Catalina Kovacs · 19 October 2026 · 8 min read
Two systems broke on the same day. Both of them told us nothing. The bugs took minutes to fix once we could see them, and finding them took an evening, which is the whole story.
By Maria Catalina Kovacs · 14 October 2026 · 6 min read
Every test passed. The function was broken. Both of those sentences are true, and the gap between them is where our site went down.
By Maria Catalina Kovacs · 12 October 2026 · 12 min read
No content management system, no database, no login. A post has a date, and a post exists if that date is not in the future. That is the whole feature. Here is how it works, what it cannot do, and why the writing is finished long before any of this matters.
By Maria Catalina Kovacs · 9 October 2026 · 10 min read
Nobody told me my phone had an SSH client in it. I found out this afternoon, and four hours later I pushed a live website from the sofa with one tap. Three things broke on the way and none of them was the hard part.
By Maria Catalina Kovacs · 5 October 2026 · 7 min read
The man who told me I was not allowed to build things had a job title that no longer exists. I have met him in every company I have worked with since, under a different name, usually with a bigger budget.
By Maria Catalina Kovacs · 2 October 2026 · 13 min read
I have been here for four of these now. Every single time, everybody counted the jobs that were ending. Nobody counted the ones that did not have a name yet.
By Maria Catalina Kovacs · 30 September 2026 · 10 min read
People moving into technology always ask me which language they should learn first. My answer has not changed in almost thirty years, and it is not Python. I have never written it down until now.
By Maria Catalina Kovacs · 28 September 2026 · 4 min read
One of the most common errors on the web means the opposite of what it says, and the fix most people are given makes their storage public to solve a problem that was never about access.
By Maria Catalina Kovacs · 25 September 2026 · 6 min read
Every startup is told that owning your infrastructure, with security and governance done properly, is a project for later, when there is money. Here is what it actually costs, and what you get for it.
By Maria Catalina Kovacs · 23 September 2026 · 5 min read
Everybody knew the structure was missing. They talked about it for years, among themselves. Once, I tried to fix it, and every single person who had agreed with me in private went quiet.
By Maria Catalina Kovacs · 23 September 2026 · 3 min read
Every week I chose between work somebody was waiting for and work nobody was, and I made the responsible choice every single time. That is the whole mechanism.
By Maria Catalina Kovacs · 21 September 2026 · 6 min read
Every change I asked for was made correctly. What was wrong was the description of the work, twice in one afternoon, and I would not have noticed if I had not been looking at the files.
By Maria Catalina Kovacs · 19 September 2026 · 10 min read
A builder is a good place to think and a bad place to be responsible. What I found when I measured my own site, and what it cost to move it.
By Maria Catalina Kovacs · 15 September 2026 · 4 min read
Agents do not fall from the sky. Engineers build them, engineers set the boundaries, and people decide how fast to ship.
By Maria Catalina Kovacs · 14 September 2026 · 2 min read
Artificial intelligence works best when it amplifies human judgment rather than replacing it. Here is what a human-centered approach means in day-to-day software engineering.