When discussing the GTM (Go to Market) strategy for NonBioS, one of the things that repeatedly comes up is how we work with IDEs (Integrated Development Environments).
One school of thought is that we integrate with IDE's and make the existing software development process faster. This is similar to how Copilots work, right in the IDE where programmers are already comfortable.
However, there is another school of thought that we have been exploring recently: Why not replace IDEs altogether?
You see, IDEs are fundamentally visual interfaces designed to make programmers more efficient. Without an IDE, the software development process is error-prone. It is easy to make syntax errors, compile-time errors, and errors in logic. IDEs make the discovery of these errors seamless and emergent. Programmers can spot errors earlier and don't have to wait until compile time or runtime. Advanced IDEs also provide visual interfaces that allow rapid prototyping, creating boilerplates, and integrating with external tools like source code repositories.
However, using something like a NonBioS agent, does this paradigm still make sense? IDEs are fundamentally designed for humans, who work with the visual world. AIs, for the most part, are blind. AI is also, in some ways, text-native. If programming is done through an AI agent, humans are fundamentally shielded from smaller mistakes like syntax and typos. Or at least that should be the objective for any good agent.
In an agentic world, programmers shouldn't have to worry about syntax and can maybe just work with pseudocode. Maybe programmers won't even have to learn different languages like Java or C, or different development environments which enable the development of software. Perhaps in the short run, programmers can get by just knowing pseudocode. It is, after all, how we were taught programming in engineering school.
Below are some of the slides we used in the discussions. Do comment below with your thoughts.
Related Posts
