Blog

There's a pattern I keep seeing with teams that adopt LLMs: the first version works, the second version is worse, and by the third version nobody knows which prompts are in production. The problem isn't the models. It's that nobody's treating prompt systems like engineering artifacts.

The pitch for no-code is that you ship faster. That's true — until it isn't. At some point the platform becomes the bottleneck, and the migration conversation starts. I've done this migration twice now. Here's what I wish someone had told me.

Musicians don't need another tool that generates music for them. They need something that listens. That's a harder problem, and it's the one I'm working on.

LLMs are creating a new level of abstraction on top of our programming languages. But the real question isn't what AI can automate — it's whether we'll finally build social platforms that don't repeat the same structural mistakes.

LLM-integrated IDEs get backlash for pushing developers out of jobs and producing broken or poor quality code. But the real picture is more complicated — and more interesting.

AI is changing a lot in the creative industry. But I don't believe it's coming for every job — unlike what Sam Altman believes. Here's why storytelling remains a skill only you can develop.

LLMs like ChatGPT have been hyped since 2022, and even I got swept up in it for a while. Here's my take on where it's really going — and why the creative economy isn't going anywhere.