Sunday, September 27th, 2026

Three things I learned this week. Not things I thought about. Things that changed how I'll do something on Monday.

Two of them cost me money before they taught me anything, which seems to be the going rate.

An engagement ended recently and it didn't end on my terms.

I'm not going to relitigate it here, partly because that's not useful to you and partly because my read on it is still forming. What I want to give you is the thing I found when I went back through it, because that part is portable.

My first instinct was to look at the work. Did I deliver, was the quality there, where did it come apart. Normal postmortem, and I ran it, and the answers were mostly fine and not especially interesting.

Then I did the thing I almost skipped, which was to go back to the very beginning and reread the intake.

And the intake was a formality.

I have a process. I've had it for years. Questions about the business, the goals, the budget, the timeline, what they've tried. It's a decent set of questions. I ask them every time.

Here's what I found. I have never once ended an intake by saying no.

Not once. Which means the intake was never a filter. It was an onboarding call with a clipboard. I was collecting information to help me do the work, which is useful, and telling myself I was qualifying, which I wasn't, because a filter that has never rejected anything is not a filter. It's a hallway.

The tell was sitting right there in the notes from the first call. There were two things in that transcript that I flagged at the time, in writing, in my own words, and then proceeded past. One was about how the last three people in my role had lasted under a year. The other was a sentence about decision-making that I remember thinking was a little strange.

I wrote both of them down. I did nothing with either.

So the lesson isn't that I need better questions. The questions were fine. The questions caught it. The lesson is that I had no step after the questions. No place where the answers got weighed against anything. No predetermined set of conditions that, if present, mean I walk.

This week I wrote those conditions down. Five of them. They're specific and a couple of them are uncomfortable because they describe situations I've said yes to before, including recently, including when I needed the money.

I'll give you three of the five, because the shape is more useful than the list.

If the person I'd report to is not the person who decides whether the work continues, that's a condition. That was the one sitting in the transcript and I read straight past it.

If nobody can tell me what the last person in this seat was measured on, that's a condition. Not what they did. What they were graded on. If nobody knows, there was no number, and if there was no number then the evaluation was a feeling, and feelings about you are not something you can work against.

If the timeline for results is shorter than the cycle time of the thing they're asking me to fix, that's a condition. This one is just arithmetic and it's the one I've overridden most often, always with some version of "we'll get early wins," which is a sentence I should probably retire.

The other two are more personal and I'm still deciding whether they're standards or scar tissue.

I don't know yet whether I'll hold the line when one of them shows up during a thin month. That's an honest thing to admit and I'd rather say it now than pretend.

But I know the difference between having a standard and having a form, and until Tuesday I had a form.

This one's small and boring and I think it's the most useful thing in this letter.

I went through my last four proposals looking for something unrelated and noticed that all four of them use some version of the phrase "includes ongoing support" or "includes revisions."

Includes. No number. No boundary. No definition anywhere in the document of what that word covers.

I did the arithmetic on what that vagueness actually costs, and the number is roughly six hours a month. Not in any one place. It's the half hour here, the quick call there, the "can you just take a look at this" that turns into forty minutes, the round of revisions that is technically the third round but nobody's counting because nobody ever said what counted.

Six hours a month is seventy-two hours a year. Two working weeks, given away, not out of generosity but out of imprecision.

And here's the part that stung. Every single one of those interactions was a good one. Nobody was taking advantage of me. Nobody was pushing. The clients were doing exactly what the document told them they could do, which was anything, because I wrote a word with no edges and handed it to them.

I've been telling myself that story for a while as generosity. Going above and beyond. Being the person who picks up. And there's some truth in it, but there's more truth in the fact that defining it would have required a conversation I didn't want to have, and vagueness was the cheaper option at signature and the expensive one for the following eleven months.

The fix took me about forty minutes on Thursday.

Every instance of "includes ongoing support" is now a number. Two thirty-minute calls per month, scheduled, within the engagement window. "Includes revisions" is now two rounds, with a definition of what constitutes a round, and a stated rate for a third.

That's it. No new document. No new system. I edited a template.

Then I did the harder half, which was the four live engagements that are still running on the old language.

I didn't renegotiate them. That would be changing the deal mid-flight and I'd deserve whatever I got. What I did was send a short note to each one saying here's how I'm defining support going forward so we're both working from the same page, two calls a month, here's the link to book them, anything urgent still gets a reply.

Three said some version of great, that's clearer. One didn't respond, which I'm reading as fine.

Nobody was upset. Nobody mentioned it again. The conversation I'd been avoiding for eleven months took four minutes to write and generated zero friction, which is roughly the same lesson as the third thing on this list and I'm going to stop pretending that's a coincidence.

The vagueness was never protecting the relationship. It was protecting me from a four-minute email.

What I'm actually sitting with is that I've been doing this for a very long time and that word has been in my proposals for most of it. I don't want to calculate the cumulative number. I ran the twelve-month version and that was sufficiently instructive.

Go look at your own proposal. Find the words that don't have edges. There are more than you think and each one is a standing invitation you wrote yourself.

There was a thing I needed to know and I'd been not asking for about twelve weeks.

I'm being vague on the specifics on purpose because the specifics aren't the lesson and because it involves somebody else. The shape of it is this: there was a question where I was fairly sure the answer was going to be bad, and where the asking itself would signal something about me that I didn't love, and so I built an impressive amount of infrastructure around not asking it.

I gathered evidence. I ran scenarios. I had a version of the conversation in my head probably forty times, which is roughly forty more than I had it out loud.

I asked on Tuesday.

The answer took under a minute. It was not the bad answer. It was not the good answer either. It was a third thing that I had genuinely not considered in any of the forty rehearsals, and it made about half of the last three months of my planning irrelevant, and it was completely undramatic.

The person on the other end had no idea it had been a question. To them it was Tuesday.

So here's what I actually learned, and it's not "just ask," because I've heard that and it didn't move me and it won't move you.

It's that I'd been pricing the wrong thing.

I'd been estimating the cost of the answer. Bad answer, this happens. Worse answer, that happens. And all of that estimating felt like preparation, like responsible work, like I was being thorough.

But the answer was never the expensive part. The answer took sixty seconds. The expensive part was the twelve weeks of running a business against a variable I could have resolved on any given Tuesday, and the specific quality of attention that gets consumed by an open question you're actively not closing. It's not a lot of attention at any one moment. It's a small tax collected continuously, which is the worst kind, because you never notice the withdrawal.

And the reason I didn't ask wasn't fear of the answer. I want to be precise here, because I told myself it was. It was that as long as I hadn't asked, all the outcomes were still technically live. Asking collapses it to one. Not asking preserves the whole set.

That's not caution. That's the same optionality trap I sent you an article about two weeks ago, running on a personal question instead of a business one, and I did not notice the connection until I sat down to write this.

So the practical thing I took from it, and the only part of this that's a system: when I catch myself gathering more evidence about a question I could just ask, that's now the signal. Not a reason to gather better evidence. The signal that I'm avoiding a sentence.

Rehearsal is the tell. If I've run the conversation more than twice in my head, I'm not preparing, I'm stalling, and every additional rehearsal makes the asking feel more consequential than it is. Forty rehearsals turned a sixty-second exchange into an event.

There's one other question I've been sitting on. I know what it is. I'm going to ask it this week and I'm telling you here so that I have to.

That's three. See you Tuesday.

One step, one day. Grace over guilt. — Dan Kaufman