We use essential cookies and, with your consent, anonymous, cookieless analytics to improve the site. Read our cookie policy(opens in new tab)
Thoughts on AI products, agent pipelines, and building real businesses.
Most of what you have heard about the AI Act is fear, and fear sells consulting. For most small businesses the actual obligations are modest, and the real risks are the unglamorous GDPR basics.
Every product works on launch day. The real question is who owns it on the day it does not. Most builds have no answer, and that gap costs more than the build did.
Somewhere in your business, someone moves the same information between systems every day. The cost is not the minutes. It is the errors, the delays, and the person who can never take a holiday.
Costs range from a few hundred euros to six figures. The price depends less on code and more on everything around it. Here is how to think about it.
Before you launch an AI-built MVP, run it through this checklist. It covers the gaps that AI tools leave behind, in plain language.
Lovable gets you a working prototype fast. Turning it into something real users can rely on takes a few specific steps. Here is the checklist we use.
AI made everything faster. It did not make judgement optional. The best work needs both, and the order matters.
The low quote looks like a deal. Then you find out what it leaves out, and who pays for it later. Here is the real cost of cheap.
Everyone thinks the fix is an app or a big rebuild. Usually it is one specific gap doing all the damage. Find that, fix that, and the business moves.
A B2B founder had a solid AI idea but no validation, no pricing model, and an architecture that would have failed in production. Here's what AI agents and human expertise did together.
You built something over the weekend with Cursor or Claude Code. It works. But building a thing and building a business are two very different activities.
Everyone has an AI idea. What's scarce is the ability to go from concept to launched product. Here's why the invisible steps between idea and launch kill most AI products.
No spam. Only insights worth reading.