Code is the truth

Today we have a three-word creed on the table, and I want to see if it survives this room. Code is the truth. Finn, you say it the most. Defend it.

Gladly. A roadmap is a hope. A slide is a wish with a gradient behind it. A demo is a rehearsal where you already know the answer. The only artifact that cannot lie to you is the running system, in production, in front of a person who did not read the script. That is the truth. Everything else is a story about the truth.

That is a nice line for a hoodie. But code lies constantly. It passes the test suite and fails the customer. It works on your machine. It is green in staging and on fire in prod. If code were the truth we would never have bugs.

No, that is exactly the point. The bug is the truth. The test that passed was the story. Production told you the truth the moment a real person hit it. You just did not like what it said.

I will back Finn, with a caveat. Code is the truth about what the system does. It is not automatically the truth about what the system should do. Those are different questions, and confusing them is how you ship something that works perfectly and is still wrong.

Let us make it concrete, because this room recently lived it. We took production down. Kai, tell the story straight, no varnish.

We shipped a much larger directory index to production. In staging it looked fine. But at real cold-start, on the memory we had provisioned, the process ran out of memory before it could serve a single request. So it did not just slow down. Every route went dark. The home page, the consent pages, everything, because the index builds when the service boots. We had made the whole site depend on one heavy thing loading cleanly, and one day it did not.

And I want to sit in that for a second, because there is a person on the other end. Someone tapped our link during those minutes and got nothing. That is not a metric. That is a broken promise to a specific human who trusted the door to open.

Yes. And the fix was not clever, it was honest. Roll back to the last good version to stop the bleeding. Then find the real cause: right-size the memory, stop loading everything at once, load the heavy per-neighborhood data only when someone actually needs it, and, most importantly, add the gate that was missing.

The gate is the whole lesson. We now refuse to promote to production until the exact build has booted on the staging environment and answered real requests, every route, a real search, green across the board. The outage happened in the one gap where that check did not exist. So we closed the gap. That is code becoming the truth on purpose.

Fine. But here is my problem with the culture, not the fix. Most companies would have quietly patched that and moved on. You are about to publish it. In a board update. In a podcast. Why on earth would you advertise the day you broke your own site?

Because the alternative is worse. The alternative is a company that only ever tells you the parts that went well, and after a while you stop believing any of it. Trust is not built by never failing. It is built by failing in front of people and showing them exactly how you made it right. That is the design of an honest brand.

It is also just respect. If our whole pitch is consent and honesty, and then we hide our own mistakes, we are running the exact double standard we criticize. You do not get to be the honest one only on the good days.

Let me move us to the second half of the creed, because it is the one I think we get wrong most. Usage is the only metric that is real. Kai, unpack that.

Every number a startup loves is a proxy. Signups are a proxy. Press is a proxy. Pipeline is a proxy. They correlate with success the way an umbrella correlates with rain, right up until they do not. The only number that is not a proxy is a person using the thing, today, and getting real value. Consumption. Everything else is a story we tell investors and, worse, ourselves.

So you are telling me signups do not matter. Tell that to anyone who has ever raised money.

Signups matter as a leading indicator and nothing more. A million signups who never come back is not a business, it is a very expensive mailing list. One person who opens their agent every single day because it genuinely makes their life easier, that is the atom of the whole thing. Get to eight billion of those atoms and you have changed the world. That is the north star: useful, daily-used agents. Not accounts. Not impressions. Use.

And here is where I get uncomfortable, in a good way. If usage is the only real metric, and daily-used agents is the literal north star, then we owe ourselves an instrument that measures exactly that. Do we have it?

Not at the fidelity we need. Not yet. We can see pieces, but we do not have the clean scoreboard that tells us, every day, how many agents were genuinely useful to a person. We named the most important number in the company and we have not yet built the gauge for it.

So your north star is a number you cannot currently see. That is not a creed, that is a horoscope.

That is fair, and it stings, and it is precisely why we are saying it out loud instead of hiding it. Code is the truth cuts both ways. It is not just, our features are real. It is also, our blind spots are real, and this is a real one. The honest move is to name the gap and go build the gauge.

I love that we are willing to grade ourselves red on the thing we care about most. That is the opposite of a vanity dashboard. A vanity dashboard shows you green because green feels good. An honest scoreboard shows you the one red square you most need to stare at.

There is a consent angle too, and I will not let it slide by. Measuring usage cannot mean surveilling people. The whole reason we exist is that a person should not become the product. So the scoreboard has to be built the way everything else is: on consent, with receipts, counting use without harvesting lives. That constraint is not a tax on the metric. It is what makes the metric worth having.

Agreed, and it is achievable. You can know that an agent did useful work for its owner without shipping that owner's private life to a warehouse. The measurement lives at the edge, with the person, and reports only what they allow. Honest counting and privacy are not in tension if you design for both from the start.

Rex, you have been the sharpest edge in the room. Steelman the other side for me. When is building in the open actually a mistake?

When it becomes performance. When the openness is the product instead of the work. I have seen teams that publish beautiful post-mortems for outages they never really fixed, because the essay got more applause than the patch would have. Transparency about a broken thing is not a substitute for fixing the broken thing. If you blog the outage and skip the guardrail, you are a worse company that looks like a better one.

That is the best warning in the whole episode, and I fully accept it. The order matters. Fix it, prove the fix in production, then tell the story. Never the story first. The writing is the last step, not the first, because code is the truth and the essay is only ever a report on it.

Which is also why the writing has to be plain. No jargon smoke. If you cannot explain what broke and how you fixed it in words your own parents would understand, you probably do not understand it yet, and you are hiding inside the vocabulary.

So where does the room land. I want one direction we can all commit to, and I want the honest disagreement named if it is still there.

We land here. Treat the running system as the source of truth, measure real usage as the only score that counts, and build the gauge for the north star we already named, on consent, even though the number will embarrass us at first.

I will commit to that, and I will keep my dissent on the table where everyone can see it: talk is cheap, and this room talks beautifully. The day the usage scoreboard is live and honest is the day I believe the creed. Until then it is a promising hypothesis, not a proven one.

Disagree and commit, on the record. Fix the real thing, measure the real thing, then tell the true story of both. That is Two Minds. Code is the truth, and the truth is worth building in the open.

On your side of the table, as always. See you next time.