Tiny Team x AI — Prototype to Production

We got our bosses to say yes to our idea… thanks to a quick and dirty AI prototype! Yesterday, shared with the STACK community our roller coaster ride of using AI coding tools….

How we got things wrong, course corrected and ended up with “passable code” from a state of “I have no idea how this AI code came to be”.

A sneak peek:

• Using a strongly typed language as it provides deep context into data structures; avoiding type issues at runtime.

• Setting up tooling for linting, formatting, testing and type checking code. Static code analysis for automatically guiding the AI.

• Balancing the use of AI vs Code Generators (i.e, hygen) or boilerplates to generate code.

• Also, flowing between the full spectrum of control through prompting styles… granting autonomy and stating boundaries to AI models.

• Providing tools like grep, rg, jq to read large documents without polluting the context too much…

• Generating documentation as you code, as context for humans and AI.

Thanks to everyone who joined and supported.

📸 Alex Chng

#InsideGovTech