Why Articles 8 and 9 are blank: and what should fill them

If you've followed this series, you've noticed something. The articles I've written are numbered 1, 2, 3, 4, 5, 6, 7, 10. There are two missing numbers. Article 8 has never been published. Neither has Article 9. They aren't drafts I haven't gotten around to. They're slots I've kept deliberately empty for almost two years.

This is the closer for the series, so I owe you the explanation. Why the gaps. What I think they're reserving room for. Why I'm in no hurry to fill them.

The short version: a constitution that's complete is a constitution that's stopped growing. The two blanks are me admitting, in print, that I haven't earned all the rules yet, and that some rules are best written down only after you've lived them long enough to know what they actually say.

What the blanks aren't

I want to clear two misreadings out of the way, because I've heard both of them.

The blanks aren't placeholders I forgot to delete. The numbering is on purpose. When I put the series together I knew Article 10 needed to be Article 10, it's the backstop, and the backstop should be at the end. The articles between 7 and 10 had to count for something even when they were absent. Empty slots in a numbered list are louder than absent ones. The gap is the message.

The blanks also aren't a coyness about rules I have but haven't published. I do not have Article 8 sitting in a private file waiting for the right moment. I genuinely haven't written them. The cursor is blinking on those numbers in the same way it's blinking on whatever rule I haven't formed yet. The unwrittenness is the actual state.

The two blanks are the bit of the constitution that says: I'm not done. Neither is the constitution. Neither, I suspect, will the constitution ever be.

What "yet to be earned" means in a personal constitution

Every rule in this series came from somewhere specific. Article 1 (no issue, no code) came from a decade of half-remembered Slack threads and reconstructing intent from diffs. Article 3 (double check, never delete) came from a Tuesday in 2017 I would prefer not to relive. Article 4 (fail fast, three strikes) came from counting, honestly, how many of my own debugging sessions past the third attempt had ever produced the fix. Each rule was lived before it was written, and the writing was the part where I admitted the lesson had stuck.

That's the bar for entry into this constitution. Not "I think this is right." Not "I read about this and it sounds smart." I have lived this enough times that I know what shape it takes when it goes wrong, and I know the cost of skipping the discipline.

The bar is high on purpose. The constitution gets its power from the fact that every rule in it has a scar attached. If I started letting in rules I merely suspected were good ideas, the document would dilute, and the rules I had earned would lose the company of the rules around them. Diluted rules don't survive contact with pressure. Lived rules sometimes do.

So the slots stay open. I have suspicions about Articles 8 and 9. I don't have scars yet. When the suspicion turns into a scar, the slot fills itself. Not before.

What I think Article 8 might become

Article 8, if I had to guess today, is something about the difference between a system I can reason about and a system I can only operate. Or (closer) something about not building the second when I could have built the first.

I keep reaching for a rule like prefer the system you can hold in your head. The pattern that produced the suspicion: I build a thing. The thing works. The thing is a little more clever than the version of it I could have fit in my head all at once. Six months later, I am back at the thing, debugging a regression, and the cleverness is what's in the way. The bit of the system I cannot hold in my head is exactly the bit producing the bug, and my only available move is to operate on it (poke it, observe it, retreat) instead of reason about it.

The cost of an operate-only system is that the model lives outside me. I can't predict how it will fail. I can't extend it without producing surprises. I can't onboard another engineer or another agent onto it without a long apprenticeship to the surprises. The cleverness was a one-time win and a permanent tax.

The rule I keep almost writing is if you can't hold the whole thing in your head, the thing is wrong. But that's not quite right, because some systems are legitimately too large to hold whole. Closer, maybe, is every layer should be small enough to reason about, and the layers should compose without surprise. That's a paragraph, not an article. The articles in this series are sentences. Eight is not a sentence yet.

The honest reason it's still suspicion and not yet scar: the shape of a rule about reason-about-able systems is going to be different in a world where part of the system's reasoning is outsourced to a model whose internals I can't read. The rule I'd write today is the rule for the world before agents. The rule I need is the one for the world after, and I haven't lived enough of the after to know what it says.

What I think Article 9 might become

Article 9 is harder to guess at, because I have two suspicions and they haven't resolved into one.

The strongest is something about time as a first-class concern. Most of my rules are about state, what changed, what's tracked, what's reversible. Almost none of them are about time. When something happened. How fresh the evidence is. Whether the model I'm operating on is from this morning or from last quarter. The bug I am most consistently bad at catching is the one where the data is correct and the staleness of the data is wrong. The cache that returned the right answer for last week. The agent's context that was accurate when the session started and isn't anymore. The runbook that was right when it was written and is now describing a system that no longer exists.

If Article 9 is about time, it might be every fact has a freshness. Or the truth has a clock; respect it. Neither is a sentence yet.

The other suspicion crowding the slot is about the cost of irreversibility, considered separately from the cost of being wrong. Article 3 covers some of this at the moment of the operation. There's a longer-horizon version I keep almost writing (build for reversibility before you build for performance) that's next-door to the freshness rule but not the same rule. I haven't decided whether they're two articles or one. The slot is open until the framing arrives.

I am suspicious of myself for having two candidates. The articles that earned their numbers came in clearly, one rule, one sentence, one shape. When the candidate is two rules in a trench coat. That's the signal it isn't ready. The slot is doing its job by refusing to fill itself.

Why I'm not in a hurry

The temptation, when you have a numbered series with two visible gaps, is to fill them for symmetry. I would be lying if I said I haven't felt it. Articles 1–10 is nicer than Articles 1–7, 10. The series would look cleaner finished.

I'm not going to fill them for the symmetry, because the symmetry is the wrong reason. The series isn't a publication; it's a record of what I've actually internalized. Filling Article 8 with a rule I half-believe to make the spacing nicer would corrupt the document in exactly the way Article 6 and Article 10 were written to prevent. The constitution gets its weight from the fact that nothing in it is performative. The blanks are the proof.

There is a quieter reason too. I want the constitution to keep growing because I want to keep growing. A document that's done describes a person who's stopped learning. The two open slots are the structural version of I expect to find more rules I need. They aren't a to-do list. They're a commitment to leaving room for the version of me three years from now to put something there that the version of me today couldn't have written.

Maybe the unwritten article is the rule about leaving slots unwritten. Reserve room for what you haven't learned yet. I keep almost writing that as Article 11 and then catching myself, the meta-rule about the gaps belongs in the gaps, not next to them. The blanks make the point better than a paragraph about the blanks would.

What this closes

This is the last piece of the Articles series and the last piece of the 51-article slate this year. It is also, I notice as I write it, the quietest. The earlier articles in this series were declarative, do this, don't do that, here is the rule, here is why. This one is the bookend, and the bookend is allowed a different register.

The register I want to leave the series in is the one where I admit I don't have all the rules. I have ten numbered slots and seven full sentences and three things that are either gaps or backstops. That's the actual state of my coding constitution at the end of 2026. Anything tidier than that would be a lie, and the constitution doesn't survive lies.

If you've read all eight published articles, you have the rules I currently know how to write down. Issue first. PR-only. Double check, never delete. Three strikes, then re-plan. Track everything. The override protocol. Test everything. Absolute obedience to the constitution itself. Eight rules, two blanks, one closing piece that says: the blanks are the point.

The next thing that fills Article 8 will arrive when it arrives. Same for Article 9. I'll know it's ready when the candidate sentence has stopped being two sentences in a trench coat, and when the cost of skipping it has shown up in my own work enough times that I can describe the failure mode without reaching for hypotheticals.

Until then, the slots stay open. The constitution stays unfinished. That's how I want it. That's how, I think, a personal constitution is supposed to be.

, Sid