Somtech

I have a long background as a software engineer and have worked at a very high level at companies such as Google and Zynga. As many of you know, I've been quite skeptical of the abilities of AI for coding. I think that stems from being used to a very high code standard. When you've spent years working with strict code reviews, you develop a pretty good eye for the difference between code that works and code that's actually good. However, I feel like things have changed quite a bit over the past six months. For simpler projects, such as dashboards, websites, and even basic games, the generated code is now good enough if you use the frontier models. I even vibe-coded an entire office suite over a weekend (check out wonky.work), and it actually doesn't suck. Sure, there are many places that can be improved, and that will take some time. But I could probably build a fully featured suite as a single developer in a month. And it's not just about building simple things quickly. For Serverpod's offline-first sync, we are doing optimizations we could never have done by hand. It's all secured by extensive tests that prove correctness, so we can let the agents go wild. We truly live in the future. With this in mind, I have an observation. I think the role of the developer is changing, but it's not going away. Someone with solid knowledge must still be at the helm. Without picking the right tools for the job, making good architectural decisions, recognizing when things start to get out of hand, or addressing edge cases, the agents' results are still not great. It's clear when you compare apps created by an experienced engineer with those created by someone new to agentic coding. It may be different in non-visual areas, but getting a frontend to professional quality is still a lot of work. Getting something to work is one thing. Getting it to look and feel genuinely polished is something else entirely. For the first time in my career, I feel like my ability to build things is limited more by my imagination than by the time it takes to write the code. That's an incredibly exciting place to be.

The Call for Papers for the @FSFlutterConf is open for another two weeks. Last time we had over 4,000 attendees, making it the world’s largest Flutter conference to date. This year, we’re aiming for 10,000. Please tag anyone you would like to see as a speaker! 🙌 This is a free online conference by the community for the community, so please help us spread the word far and wide. 🤩