Week of June 15–21, 2026 — Built the Machine

The one-liner: I stopped building websites one at a time and built the machine that builds them instead.

What Happened

Monday — Ship the words

Kicked the week off by publishing three articles for the agency blog — AI search for tradies, Local Services Ads, and speed-to-lead. The kind of posts that quietly do their job in the background for years. Cover images, in-body diagrams, the lot. Felt good to start the week shipping rather than fiddling.

Tuesday — Learning to talk to a camera

Spent the day on reels research — specifically the unglamorous part nobody films: how to actually record a talking-head video that doesn't look stiff, and why my edits kept fighting me. Wrote up the recording learnings so future-me doesn't relearn the same lessons. Also sat down and thought hard about whether building this whole thing from where I am right now actually makes sense. Concluded it does.

Wednesday — Quiet

A gap day. Not much hit the keyboard. Some weeks have one of these and pretending otherwise is just theatre.

Thursday — Filling out the agency site

Built out the agency site properly — added custom software and automation as services, shipped a speed-to-lead tool, fixed the footer grid that had been bugging me, added a sitemap and a custom 404. Also wrote a local-sales playbook so the outreach has an actual spine instead of vibes.

Friday — Lead machine + the spark

Two things. First: built lead magnets — a tradie website checklist and a “get found on Google in 7 days” guide, wired to reusable opt-in pages that drop straight into the CRM. Second: scraped and enriched a few hundred local plumbers and property-maintenance businesses, sorted into tiers. Then late at night the real idea landed — instead of hand-building each client site, build one engine that generates them from a config file. Started immediately.

Saturday — The big build

This was the day. Built the tradie site engine end to end — a config-driven system where you fill in one file and a complete, polished client website falls out: hero, services, galleries, reviews carousel, suburb-level SEO pages, schema markup, a blog. Roughly a hundred commits of design polish and performance work in one sitting. Merged it into main and spun up three demo sites off the same engine to prove it works.

📷 Three different demo sites, same engine

Sunday — Publish and breathe

Lighter day. Pushed out a content piece on the AI-agency space getting flooded, and let the week settle.

What I Built

• Three SEO articles published to the agency blog (AI search, Local Services Ads, speed-to-lead)
• Custom software + automation service pages, a speed-to-lead tool, sitemap, custom 404
• Two lead magnets with reusable opt-in pages wired into the CRM
• A few hundred enriched, tiered local-business leads (plumbers + property maintenance)
• The tradie site engine — config-driven generator that produces a full client website from one file
• Three live demo sites off that engine
• A local-sales playbook and reels recording research notes

What I Broke / What Was Hard

• The mobile floating call/quote bar kept sitting too low and jumping the page when tapped — chased that one across several passes
• Service-detail pages were missing their images until I wired in a fallback to the gallery photos
• Page-speed on image-heavy pages took real work — converting trust badges to WebP and routing everything through proper image optimization
• Footer layout fought me for an embarrassing number of commits before the columns sat evenly

The Honest Take

This was a pure build week — head down, barely came up for air, and the diary stayed empty because of it. Normally I'd flag that as a red flag, but this one earned it: I turned “build a client site” from a multi-day job into a fill-in-the-config job. That's the kind of leverage I actually care about — work that makes all the future versions of the work cheaper. The trap now is admiring the machine instead of feeding it clients. An engine with no one running through it is just a nice-looking repo.

Next Week

Point the machine at a real client and the tiered leads at a real conversation. Building was the easy part — selling is the test.