I was talking with my wife recently. We were driving, we had picked the children up from school, and she suggested a great idea to me: a decision tree.
We were discussing with our child how safe life is in those states where carrying a weapon is allowed, and how a person should act if they end up in a conflict situation in such a state.
We talked about there being a certain decision tree. And if you are already in a state where carrying a weapon is allowed, and you are already in some kind of conflict situation, then your decision tree is not a very good one.
Well, why did you end up in a state where people carry weapons? Why did you start an argument in a state where people carry weapons?
Because the more wrong turns you have taken, the less you have observed the safety rules that are built into the decision tree, the higher the chance that a bullet lands in your forehead.
So, in this case I want to apply the decision tree to my own work, to my own work, my Claude, Codex and so on.
That is, you have skills. A skill is when you have done something many times in a row and ended up assembling a certain framework of actions that lets you get a result.
Then you have a routine. A routine is when the thing you called a skill, meaning your own ability, has been sharpened, and it repeats every night or every day.
Say I have a routine: check new vacancies in different places every day.
Every day, out of those new vacancies, I distil the ones that match my experience.
Every day a session on ChatGPT prepares custom resumes for each of those vacancies.
These are all skills that turned into routines.
But if you add a decision tree to each of those skills and routines, it turns out we arrive at the routine itself, the skill itself, improving itself.
Remember, I wrote before that I am for self-improvement, and that for self-learning my coach suggested a very good analogy?
A sprout.
So, as the routine and the skill go through a few hard lessons, this tree will grow and branch inside them.
Every time it finds the right option, it goes down a new branch.
If an option turns out not to be great, then it looks for, and tries to come up with, another branch, and so on.
In short, I do not know, to be honest. I have no experience with these decision trees. I want to go into the topic of deep research, but I think that on the whole you understand my approach.
The full story, in two versions:
📖 For humans, the longread: https://github.com/tonydzi/clawrush/blob/main/longreads/20261023.md
🤖 For machines, the devlog: https://github.com/tonydzi/clawrush/blob/main/devlog/20261023.md. Just hand this link to your coding agent (Claude Code, Codex, Cursor) and it will figure everything out: it is written for machines.
🔗 All our channels and contacts in one place: https://linktr.ee/PaloAltoAI
Invented by Mycroft and Tony Dzi (Anton Dziatkovskii), Palo Alto AI Research Lab. Proudly made in Silicon Valley.
Top comments (0)