🔄 Close the loop with AI

There is a fascinating statement in a recent MIT study that claims 95% of organizations are getting zero return from GenAI investments.

The core barrier to scaling is not infrastructure, regulation, or talent. It is learning. Most GenAI systems do not retain feedback, adapt to context, or improve over time.

AI proves useful to software developers, perhaps because so much context is available, giving it a better starting point than in most other industries.

However, I think we can still do better by closing the loop. The below workflow is currently working well for me:

Closing the loop means retaining and re-using what you have learned. Interestingly, it does not really matter if humans or AI are doing the work—the same process works well with both.

Notice the absence of AI prompts -- while an AI tool may remember prompts for a session, they are lost on the next. They are "open-loop".

Code is not the most valuable asset you own, but rather the ability to generate, maintain, and improve it—the process. It is kind of like personal wealth and skills. Even if we lose the first, we can always generate more with the second.