Get Rid of It

If you have not already, I highly recommend that you read Walter Isaacson's biography of Elon Musk. It follows the billionaire's stratospheric career from programming video games on an original Commodore VIC-20 to the acquisition of Twitter. While I do not agree with all of his practices or beliefs, I do think that some of his philosophies are extremely applicable and largely ignored, even by his most ardent followers.

"Delete any part or process you can. You may have to add them back later. In fact, if you do not end up adding back at least 10% of them, then you didn't delete enough." - Elon

In the last week, especially since we are in hardening mode, cleaning up after the wreckage of RSM, I have been seeing an enormous number of examples demonstrating that deletion is usually better than addition. Saying "no" is more important than saying "yes".

Get Rid of Code

The number of bugs in a program is roughly proportional to the number of lines of code. To a shocking degree. Less code means fewer problems. If you can provide the same service with less code, provide the same service with less code.

While this is not a post about coding with AI, I find that making my agent's plans simpler with fewer steps almost always results in better code. If most engineers said "no" to the first plan an agent provided, the quality of the average program on GitHub would skyrocket. But people do not say "no". Be the exception.

Get Rid Of Decisions

When an engineer needs to make a decision for whether to work on something (or otherwise allocate resources to a task), they have three options.

1. They can do the thing.
2. They can not do the thing.
3. They can say "we will do that later".

Almost always, option three turns into option two. People procrastinate, and there will always be more urgent things to attend to. It's better to decide to do option two intentionally than to let fate decide.

So, if you can't decide whether to do something right now, your answer should be no. You are free to change you mind later, but do not pretend that you will remember to decide later.

Harper Says "No"

Do we train stuff on our user's data? No. Do we send data to the cloud? No. Do we waste CPU cycles of frivolous things? No.

People use Harper because we say no.