Week of July 20–26 — Killed Three, Built One

The One-Liner

Killed three businesses on Tuesday morning and had the thing that replaces all three of them live by Tuesday night.

What Happened

Monday — Ship day

Deployed Closewright — closewright.com — a site I'd been circling for weeks. It's the sales-education thing: teaching technical founders how to sell, because the people who can build almost never can, and it's the most expensive gap I've watched good builders carry. Next.js, one job on the page: book a free teardown. Nothing clever, just built and live.

Also pushed two new posts to the agency blog and — more usefully — built a generator for the cover images so I stop hand-making them one at a time. That's the trade I keep trying to make: spend an hour building the thing that removes the recurring hour.

And a small one that isn't small. A digital product I've been quietly running made its first paid sale. Twelve dollars and fifty-seven cents, net, to someone who found it organically. That is not a business. But there's a difference between zero and not-zero that no amount of planning gets you across, and I'd been sitting on zero for months.

Tuesday — Killed three things, built one

Started the morning by killing three business ideas in a single sitting.

An ads agency in a channel I'm genuinely good at. A publishing play. A third I'd been sentimental about for over a year. All three failed the same test: distribution treadmills. Work that only pays while you're actively standing on it, in channels where the platform holds all the leverage and I hold none of it.

The distinction that made them killable — kill the business, keep the skill. The ads knowledge doesn't evaporate because I've stopped trying to sell it as a service. It just gets pointed somewhere better.

Then I spent the rest of the day pointing it somewhere better. Built and shipped ACoS IQ — acosiq.com — start to finish, live by the end of the night. Not a landing page. The whole thing: services, results, six long-form articles, three free calculators, a generator for the article cover images, email routing, legal pages, and a machine-readable summary file for AI crawlers since that's increasingly how people find anything.

It does two jobs. It's the proof page I send from outreach — here's the work, here's the evidence, judge for yourself. And it's a slow SEO layer that compounds while I'm asleep. No pipeline, no headcount, no treadmill. One asset I own.

Best day of the week by a distance, and the shape of it is the lesson: the prune wasn't the opposite of building. It was what made the building obvious. I'd been carrying three vague businesses for months and couldn't have told you what to work on. Delete all three and the actual move takes about ten seconds to identify.

Three months ago I'd have kept all three alive as "options." Optionality is mostly just indecision that's learned to dress well.

Wednesday — hole in the record

No diary entry. Whatever happened didn't survive contact with my own note-taking, which is its own finding.

Thursday — Unglamorous prerequisites

Two landing pages for the agency. Website builds, and Google Business Profile management. Boring, necessary — you can't point paid traffic at a homepage and hope it figures itself out.

Friday — Deliberate rest

Cooked. Properly cooked. I'd been pushing without a break for long enough that the tiredness had stopped being tiredness and started being something with a shape.

So I took the whole day. No applications, no work, no "light admin" that's actually work wearing a hat.

The part I want on the record: this was a decision, not a collapse. I've done the version where you grind until the body makes the call for you, and the recovery cost is roughly ten times higher. Calling it early is cheap. Calling it late isn't.

Saturday — The reassessment

I redid my own personality profile from scratch. Not a fun-quiz thing — a proper read across everything I've written about myself over the past couple of years, deliberately ignoring what I'd previously concluded.

The finding rearranged something. I've spent years describing myself as the analytical, systems-first, cold-decision type. The evidence says that's a built layer, not a native one — an operating system I constructed around 22 and have been running ever since. Underneath it is something a lot more feeling-driven than I've been willing to claim.

The two things that landed hardest:

The systems aren't personality, they're prosthetics. The build logs, the kill lists, the rules I write to stop myself doing the thing I'd otherwise do — I'd been reading those as evidence of discipline. They're the opposite. A naturally structured person doesn't need scaffolding to be structured. Mine is load-bearing. Which is fine, and worth knowing, and means "just be more disciplined" was never going to be the fix.

Structure is downstream of stability. The order is stability → security → structure → building. Every time my discipline has deteriorated it's tracked chaos in my life, not some failure of character. I'd been coaching myself as if it were a character problem. Wrong lever, for years.

The other thing that surfaced: a few years back I got deep into abstract acrylic painting. Loved it, genuinely in the zone doing it. I'd always assumed it stopped because it wasn't going anywhere. It didn't. It stopped because of the judge — an internal scoring reflex that sat between me and any feeling of having made something. Abstract work has no correct answer, so the judge imports a standard from somewhere and blocks the reward.

The skill to learn there isn't painting. It's making something without scoring it. I have almost no reps at that.

Sunday — Late night, good problem

Went down a rabbit hole on the voice agents I've been building and hit the good kind of wall — the one where you discover the thing you were about to build from scratch already exists as a setting somebody shipped.

I'd been sketching a whole custom pipeline. Turns out I can keep the voice layer exactly as it is and just point it at my own model endpoint. Days of work collapsed into an afternoon, and the version that survives is simpler than the one I'd designed.

Best possible outcome, mildly annoying delivery method: the research you skip is the research that costs you the most.

What I Built

• acosiq.com — full site built and live in a day: services, results, 6 long-form articles, 3 free calculators, cover-image generator, email routing, legal pages, AI-crawler summary file
• closewright.com — sales-education marketing site, built and deployed
• Two agency landing pages (website builds + Google Business Profile) — prerequisites for running ads
• Reusable blog cover-image generator for the agency blog — kills a recurring manual job
• Two agency blog posts published
• Rewrote my LinkedIn profile end to end
• Architecture decision on the voice agents that deleted about 80% of the planned build

What Broke / What Was Hard

• Diary gaps. Two missing days in seven. The week's record is mostly reconstructed from commits, which tells you what I shipped and nothing about what it cost.
• Badly front-loaded. Everything real shipped Monday and Tuesday. By Thursday I was on prerequisites, Friday I was flat on my back. The two-day sprint and the Friday burnout aren't separate events — one caused the other, and I'd like to stop pretending that's a coincidence.
• The burnout signal. Recognised early, which is progress. The pace pattern that produced it is unsolved, which is not.
• A site with no traffic. ACoS IQ is live and nobody has been to it. Shipping is the cheap half.

The Honest Take

Two brands shipped, three killed, one rest day, one personality rebuilt. For a week I'd have described in the moment as scrappy, that's a strange amount of ground.

Tuesday is the one to keep. Not because of what got built — because of the order. Three plausible businesses died in the morning and the right move became obvious by the afternoon. Plausible is precisely the problem: plausible ideas are the ones you can carry for years without ever having to admit they aren't happening. They don't fail loudly enough to bury. They just quietly eat the attention that the real thing needed. I'd been telling myself I was short on ideas. I was long on ideas and short on deletions.

Saturday will still matter in a year. Finding out that the structured, analytical operator I've been for most of my adult life is something I built rather than something I am isn't destabilising — it's a relief. It explains why the scaffolding has to be so elaborate. It explains why it collapses when life gets chaotic rather than when I get lazy. And it means the fix for a bad stretch is stability first, not another rule on top of the rules.

The rest day is the same finding wearing different clothes. A year ago I'd have ground through it and called it discipline.

What I'd flag against myself: shipping a whole site in a day feels excellent and is exactly the pattern that put me on the floor by Friday. I'm good at the sprint. I'm bad at the week. The prune was a real decision; the pace still isn't one.

Next Week

Ship the voice-agent architecture properly and get the daily rhythm back — the pace problem doesn't get fixed by another rule, it gets fixed by not letting the week run itself.