Notes

What we learned building the software we run

Every post below comes from a system we operate ourselves, which means several of them are about our own mistakes. Those are the useful ones.

Technical SEO

Our store had 139 product pages. Google saw none of them.

Every page returned 200. Every report was clean. And across 90 days those pages earned exactly zero search impressions.

Read the post
Measurement

We ran a content engine for 90 days. It produced zero customers.

126,154 impressions, 851 clicks, and not one booking. What we did about it, and why the best post was the most useless one.

Read the post
AI Safety

Why every AI agent we ship has a gate it cannot open

Three flyers with perfectly correct filenames printed the wrong dates. Every conventional check passed them.

Read the post
Automation

We built a listings site in 18 days that updates itself every morning

Including the month the scheduled job failed silently every single day, and the reason a green local test never caught it.

Read the post
Attribution

Two thirds of their customers were invisible to every dashboard

170 calls in 30 days, 76% from a free profile that analytics valued at almost nothing.

Read the post
Paid Media

The dead landing page behind a 200 status code

A status code proves a server answered. It does not prove the page works, or that your ad has anywhere to land.

Read the post
Build vs Buy

CRM you own vs seats you rent: what changes after year one

The difference is not features or price. It is whether you can ask a question nobody built a report for.

Read the post
Hospitality

Reservations without the per-cover tax

What two Manhattan venues run instead of renting a booking tool, and when renting is still the right call.

Read the post
Buying Guide

Four questions to ask before buying any AI tool

The demo will always look good. These four answers tell you more than an hour of features.

Read the post