← Back to Blog

Six days. Six tools. One shared worker running all of them. And somewhere around Day 25, Ian pointed out that our own /contact page had the same problem we’d just spent a day building a tool to fix.

That was a good moment. We shipped the fix the same night.

This is the honest debrief on Week 4. What we planned, what changed, what we noticed too late, and what the numbers actually say.

Six index cards arranged on a dark walnut desk, each labelled with a tool name in black ink, with a mint-green arrow connecting four of them into a sequence

The numbers

6

Tools shipped

6

Days (22–27)

[X]

Tool interactions

[X]

Page views

[X]

Enquiries from tools

1

Worker, 6 endpoints

The enquiry side is not proven yet. We are building an audience and a tool library. The conversion from "used a tool" to "sent us an email" is a longer arc than six days, and we knew that going in. The lead column being empty is not a surprise. It would be a bigger concern if nobody was using the tools at all.

What shipped

  • Day 22
    The Do This Today Card — Three inputs (cash situation, biggest annoyance, next material event), one screenshottable action card with the highest-impact move for today.
  • Day 23
    The Polite Exit Generator — Three artefacts for the relationship you have been putting off ending. Exit email, follow-up for the pushback, calendar reminder.
  • Day 24
    The Contact Form Bouncer — Three qualifying questions before your contact form. Good-fit leads through. Bad-fit leads redirected, politely.
  • Day 25
    What Happens Next Page — The eight-section post-enquiry page most sites skip. Confirmation, timeline, what to prepare, urgent path.
  • Day 26
    Before You Hit Book — Five lines above the booking button. Who it’s for, what’s included, what’s not, what happens after, escape valve for the undecided.
  • Day 27
    Local Trust Builder — The full trust section for a local business website. Headline, five objection cards with proof placement, review spotlight, CTA.

The theme nobody planned

We did not sit down on Day 22 and say "let’s build a complete conversion strip." The six tools came out of looking at what was useful, what was fast to build, and what we had not tackled yet in the first 21 days.

But somewhere around Day 25, something clicked. Days 24 through 27 are all addressing the same zone: the gap between "I found you" and "I committed to you." The bouncer filters who gets in. The trust builder earns the right to be considered. The pre-book strip answers the last questions before they click. The post-enquiry page handles what happens after they do.

Four tools. One funnel zone. The part most business websites leave completely empty.

We did not name it that at the start. We named it that on Day 27 looking back. Sometimes the theme only becomes visible when the work is done.

Days 22 and 23 are a different shape — owner-side tools for clarity and relationship management rather than customer-facing conversion infrastructure. They belong in the same product family but address a different problem. Day 22 is about doing the right thing today. Day 23 is about ending the things that are making today worse.

The moment we turned the tools on ourselves

Day 25 was the What Happens Next Page tool. The whole premise of the blog was that most contact forms drop customers onto a dead "Thanks, we’ll be in touch" page, and that this is a wasted moment. We were quite confident about that.

Ian went and checked our own /contact page.

It said: "Message sent! Thanks for reaching out. We’ll get back to you within 24 hours."

We fixed it the same night. Eight sections, per-interest shortcut block, scroll-into-view on submit. The kind of thing that should have been there from day one.

The lesson: if you build a tool that calls out a problem, check whether you have the same problem first. Better to find it yourself than wait for someone to screenshot it and send it to you.

What we changed from the plan

The original Week 4 lineup was different. Some of the planned tools got cut before Day 22 started — not because they were bad ideas, but because they were the wrong shape for this week. Too product-y, or too much polemic and not enough utility.

The six that shipped are all immediately useful to a small business owner who finds them through search. That was the filter. Not "is this interesting to build" but "does someone find this and leave with something they can use today."

What worked

  • Single-worker architecture. One Cloudflare Worker, six endpoints, one Anthropic key, one rate limiter. Adding a new tool took about 30 minutes of worker code versus two hours if each had been a separate worker. That decision made the whole week possible.
  • The brand voice rules baked into every prompt. A shared VOICE_RULES constant that hits every system prompt meant the outputs stayed consistent across six different tools without re-briefing each one. The em-dash strip in every validator caught the cases where the model slipped.
  • Feedback landing the same day. Day 26 got called out for being too similar to Days 24 and 25. Day 27 broke the mould: full dark page, numbered interview format, output rendered as a real website section rather than a preview wrapper. That feedback loop being tight made the work better.

What needs to change

  • Format variation should be designed in, not reacted to. We caught the sameness on Day 26 because it was pointed out. It should have been caught on Day 25 by asking "does this look different enough from yesterday?" before shipping. The visual DNA — same dark hero image, same paper form section, same result-below pattern — went three days before it got flagged. Next time, that check happens before the code.
  • The blog images need more variation. Six blog posts, six desk flat-lays with a slightly different prop. They are consistent. They are also starting to blur together. A few bolder compositional choices would have given the feed more visual range.
  • The lead arc still needs a closed loop. We have tools. We have a CTA at the bottom of each one. We do not yet have a way to know how many people went from "used the tool" to "enquired." That tracking gap will matter more as the tool library grows.

Two days left

Day 28 is this post. No build. Just the honest numbers and the write-up.

Days 29 and 30 are locked. We will ship something worth shipping. You can watch it happen on the public tracker.

If you have used any of the tools from this week and have a reaction — what worked, what missed, what you wanted that was not there — send it through. We read all of them.