A retro arcade shooter where Claude Code wrote all the game code and DALL-E generated every sprite. No human code written.

The weary pilgrim tosses his last coin into the well, whispering his deepest desire to the dark waters. He hopes some hidden magic might make it real. The dream of the wishing well is as old as humanity. The genie in the bottle, the magic wand, the monkey’s paw. People have always wanted to speak their desires into existence. AI code generators like Lovable or V0 are just the latest way to meet this age-old desire. ...

There’s a world of difference between Andrej Karpathy’s vibe coding and the kind you’d see from most product managers. The former understands what the machine is doing under the hood. The latter talks about specs without understanding the inner workings of the product increment. A better name for that workflow would be “blind vibe coding”. It’s hard to overstate just how different non-techies talk to their Cursor agent. From the original prompt to the intermediate feedback, a technical background makes everything easier, more informed and better. It’s tough to put myself in the shoes of those who lack those technical chops. ...

AI tools are overhyped, but their results undeniable. They turbocharge software engineers. Product managers, on the other hand, struggle to find the same productivity boost. Sure, they can create an interactive mockup in Lovable, but that’s of very limited use on brownfield products. So what can a PM do? There is a tempting, popular idea out there that leads to a dead-end street. The cardinal sin of AI-powered product management is having Claude generate your Product Requirement Documents for you. It’s unfortunate how popular that sin is getting. ...

An experiment in using AI code agents to write all the missing unit tests for an existing codebase. Let’s take code coverage to 100%.

Rock climbing is one of my favourite ways to blow off steam. Whenever I tell people about that hobby, I get a common response: “Sounds really cool, but I could never do that because of my fear of heights…” Fear of heights is real, and it’s something we all have. In order to climb, we need to conquer that fear. And the only way to conquer that fear is… to climb. ...

In my line of work, I see a lot of legacy code – products that were built so fast the foundations are weak. Every little change causes something else to break. The industry standard way to stabilise such a codebase has always been test automation. But writing unit tests takes time, and in the rush of shipping features, that time is often dedicated to anything but tests. As the product expands and the code gets ever more complex, this lack of tests becomes a painful problem. If small changes break the system, sweeping changes are downright scary. The team gets paralysed. ...

LLMs are productivity multipliers for software engineers. Over the last few years, the quality of their output has reached a professional level. They cannot one-shot a complete SaaS product (yet), but it’s more than just fancy autocomplete. What used to take an hour now takes minutes. Once you get used to building with AI-powered tools, the alternative feels awkward. I remember developing software in the early 2000s. The internet existed, but treasure troves like Stack Overflow didn’t. Building something with a new framework meant buying and reading the physical book. Before hoodies were in, the sign of the real developer was a stack of O’Reilly books on your desk. ...

People, in general, don’t care about privacy. They like the idea of their data being safe, but their actions tell another story. They’ll click on any link and sign up for the shadiest free tools without giving it a second thought. If you’re not paying for a service, it means you’re the product, and people have shown over and over again that they are OK with that. AI tools take this careless behaviour to the next level. I think it’s safe to assume that ChatGPT hears more confessions than the average priest. People freely share medical information and financials with their LLM. ...

LLMs are stochastic parrots. They just predict the next most likely word. And the most likely token to follow “AI” these days is, without a doubt, “Agents”. AI agents are hailed as the next quantum leap in the evolution towards building SkyNet AGI. An agent is basically a program that can interpret input with an LLM and has access to tools to actually do something with that input. Instead of ChatGPT drafting you an email, an agent could send emails on your behalf. You could ask your agent to “send friendly reminders to fill out that Doodle”, and it could read your latest Doodle, see who didn’t reply yet and then ask the Gmail agent to send those reminders to the right people. Rather than text-in-text-out, an agent converts input into actions. ...