The message of this post is exceedingly simple: don't build a system or a habit too early, do it by hands first until the dust has settled. Nothing especially insightful.

But it needs reminding if you happen to dabble in experiments on a frequent basis.

Premature systemization is corollary to premature optimization and do things that don't scale.

Let's demonstrate. You and teammates think more sync up is needed but not sure what's the best way to do that. You start by scheduling a weekly meeting to solve this problem, not knowing if it will work.

Three months later, this meeting still happens weekly but the problem of sync up remains unsolved (somehow). Now the team is additionally burdened with running another meeting that solves nothing.

Another example. You have a health problem, so you decide to take a specific supplement. Knowing it will take multiple months before it shows effect, you kept taking this supplement for years not knowing if it works.

Systems and habits are in place as energy conservation devices. It's great when they work, but only insofar as they are effective. These systems crept in and got stuck.

You can't know that without periodic purge-review, as process of asking "what system/habit needs to go."

Many things are supposed to be experiments. Experiments are expected to fail; failure should be a positive call for change.

The longer a habit sticks, the harder is it to shake. It's hard for a person, even more so for a group.

A system is prematurely installed when it makes wrongful assumptions on established patterns, or even patterns that are not established at all. It's like placing a trend-based bet based on a sample of one.

Some reasons why this mistakes get made:

• Wishful thinking about having found the pattern (one sample doesn't make a trend)
• Wrong extrapolation of history
• The field is still in flux
• Playing with dynamic situation is slow, cumbersome and costly

The current pace of generative AI reeks of a playing field in flux. It's changing fast enough that most responses to it will be obsolete in three months.

Juggling the timing for sysmtemization might be something uniquely for humans.

Systems belong in the realm of order. Learning & discovery belong in the realm of chaos. The game is in timing them one step ahead.