> ## Content Index
> Fetch the complete content index at: https://www.taivo.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# Close loops
- URL: https://www.taivo.ai/close-loops/
- Published: 2026-08-23T09:36:19.000Z
- Updated: 2026-08-23T20:00:28.000Z
- Description: When the right action is unclear, progress comes from a simple loop: act, observe, learn—and change what you do next.
- Author: Taivo Pungas
- Tags: Work and judgment, stream

Doing something new is hard. But it is hard not in the usual way, where the right action is painful or difficult. It's hard because the right action is unclear. Figuring that out requires trying many things to learn what the right thing to do is in the first place.

Here are a few examples of when the path is essentially known, just hard to follow:

- Learning to play the piano at an elite level of performance
- Hitting a specific body fat percentage

Contrast it with the hard-because-unclear kind:

- Getting users to stick around with a product
- Making a robot operate without human input

For the first kind of hard, you mostly need grit. But for the second kind, you need to *close your loops*. A loop at its minimum has three steps:

1. Act
2. Observe
3. Learn

It is simple to the point of triviality. Yet stopping at any of the three steps creates a distinct failure mode:

1. Failure to act: theorizing in an armchair.
2. Failure to observe: hustling without any awareness.
3. Failure to learn: when your next action is not a function of the previous loop.

Among smart people I've worked with and reflecting on my own work, failure to learn is the most common one. It doesn't take much to act, and "metrics" and "KPIs" are easy to set up in a technical sense. But truly closing the loop is rare. When it happens, it's magical.

If what you learned doesn’t change your next action, the loop is still open.

*This is an age-old idea with many variants. A few related frameworks:*

- *Scientific method*
- *OODA loop (observe-orient-decide-act)*
- *Lean startup (build-measure-learn)*
- *The 2026 discussion around self-improving companies (*[*YC lore*](https://www.taivo.ai/stream/blomfield-loop)*)*