Learning, as well as code development is boosted when you can minimize the time between getting feedback and applying what you’ve learned. Iterate on small changes and assimilate knowledge in between rather than ingesting a ton of information without reflection, or making big decisions based on incomplete information.

Useful also for product development (Lean Startup) to see if a proposed business model is viable.

Similar in idea to the Japanese term “Kaizen” for continual improvement in working practices and efficiency.

Can a feedback loop be too short?

If Long Feedback Loops Inhibit Improvement, is the other extreme also something to avoid? In the world of anomaly detection, a certain amount of history is required in order to see patterns in the noise, detect cyclical events, make predictions, etc. So maybe extremely short feedback loops (Polling Frequency) aren’t an issue, but you still need to have the historical context to make sense of the data.