A bit over a month of intense experimenting with LLM-based programming tools.
The results are very convincing — in small projects, most tasks can be greatly simplified.
Things like documentation lookup, prototyping, integrating with unknown APIs, implementing different architectures, and transforming between architectures and tools — all of that works across all programming languages.
Documentation writing, security analysis, and hardening are also well supported.
What still takes the same amount of manual work is fine-tuning the GUI and improving usability — that part can’t really be automated yet.
We’ll see how it goes. We’ve already completed over a dozen test projects.
