This rule is the cure for the pattern that almost cost me a decade. I would watch, read, talk, and somehow leave no record of any of it. Interest without an artifact is not the same as understanding. Understanding without an artifact is not the same as competence.

The artifact does not have to be good. It has to exist.

A small artifact compounds. A streak of small artifacts becomes a body of work. A body of work is the only honest claim that learning happened.

This identity is not aspirational. It is operational.

Every domain in my life has a doing form. Software has code. AI has agents and evaluations. Math has solved problems and proofs. Physics has derivations and simulations. Electronics has breadboards and PCBs. Philosophy has essays and argument maps. If I cannot point at a thing I have produced in a domain, I have not yet practiced it.

The reverse is also a rule. I do not build without studying the canonical source. I do not ship without testing. I do not publish without revising. Builder and researcher and engineer are three angles on one practice, not three jobs.

The longer version of this principle is published as the The AI Usage Constitution in the notebook. The short version: AI compresses time, but only against knowledge. Against no knowledge, AI does not compress, it just produces, and the producer has not grown.

I refuse to be the producer who has not grown. The fastest path to fake competence is also the fastest path to no competence. The detection mechanism is simple: can I explain the code, the derivation, the report, the design, without the AI in the room. If yes, the tool worked. If no, the tool worked against me.

This principle is here because I have a tendency to either burn through something at unsustainable speed and abandon it, or to wait for the perfect moment and never start.

The middle path is the one I am training. Work that is consistent and never urgent. A Tuesday hour. A Saturday hour. The month between them is not the point. The five years they accumulate into is.

This sits in deliberate tension with the-work-is-temporary. The resolution is that urgency belongs to the year, not to the morning.

I have spent too many hours watching summaries of things I could have read in the same amount of time. Summaries are a translation. Every translation loses something.

A YouTube video about a paper is not the paper. A blog post about Linux internals is not the kernel source. A tutorial on calculus is not the textbook by the mathematician who taught the course at MIT.

The rule is not that summaries are forbidden. The rule is that summaries are scaffolding, and scaffolding comes down. If I am still on the scaffold a year later, I have not built anything.

The reviews are on a schedule because review is what makes the public log useful instead of performative.

• Sunday: weekly accountability note.
• End of month: site-state notes.
• End of quarter: goal review.
• End of year: full retrospective.

The point of publishing the work is not the audience. The point is the accountability that the audience produces. The audience is the side effect that keeps the work honest.

The other discipline is response: I read what comes back. I let it change positions where the argument is good. The changed-my-mind log is the artifact of that.

This principle sits in deliberate tension with no-artificial-time-pressure. The resolution is that urgency belongs to the decade, not to the morning. I refuse to ruin the morning with manufactured panic. I also refuse to live as if the years are infinite.

The artifact rule is partly a response to this. The first 90 artifacts in the master plan are partly a response to this. The choice to publish before the work is perfect is also a response to this.

The work that lasts is the work that compounds. The work that compounds is built one boring day at a time. The point of admitting the temporariness is not to rush. The point is to make the boring days count.

memento mori.