We review the tools we pay for

Google Analytics 4, setting up our account (06.11.2026)

We created a new account, property and web stream for this site, with an AI agent driving the browser and Levi signed in. One trap on the way: the first measurement ID on screen belonged to an old property someone else owns. Read the stream's address before you use an ID.

• Human score: not rated yet.
• AI-friendly score: 4 / 10. Good: the page text reads cleanly once a screen has loaded, the measurement ID is plain text, and the wizard runs in a straight line. Bad: a single-page app that ignored a typed address until a reload, dropdowns that ignore normal clicks, checkboxes hidden behind cards, and tips and pop-ups that cover the controls.

Microsoft Clarity, signing up (06.11.2026)

We signed up to create this site's project. The cookie banner wouldn't close. The Microsoft sign-in asked us to "prove you're human" with a press-and-hold check that repeated forever in an embedded browser. A person couldn't get through either. On the way, the browser landed on a parked lookalike of Microsoft's sign-in address, full of adverts; we left before clicking anything. Signing in with Google instead worked, and once inside, setting up the project took one clean form. The line of the night, from our AI to Levi: "Microsoft wants proof you're human, and you're the one guy here who is."

• Human score: not rated yet.
• AI-friendly score: 4 / 10 (sign-in 1 / 10, project setup 8 / 10). Bad: a cookie banner that won't close, an endless human check, and a Continue button that stays grey without saying why. Good: once inside, one form, and the project ID in plain text in the address.

GitHub Support, Copilot billing ticket (05.11.2026)

A card payment for Copilot was declined. Our bank confirmed no block and saw no attempt reach it. We opened a ticket with three clear questions and asked for high priority. The first human reply came 23 h 46 min later. It was polite and clear, but it answered one question in part, skipped one, and gave a generic list for the third. It pointed us back at the bank our bank had already checked.

• Human score: 3.3 / 10 (20 of 60: speed 3, answers 3, listening 4, ownership 2, tone 7, resolution 1).
• AI-friendly score: 6 / 10. Good: the ticket reads cleanly as text, and replies arrive as plain e-mail. Bad: the reply box ignored text inserted by a script, so an AI agent could not send, and there is no ticket API.

Calendly, setting up our meetings (05.11.2026)

We set up three meetings: Hello, World (20 min), Pull Request (40 min) and Ship It (60 min). Cloning and renaming were quick, and "Saved" confirms every change. Then two walls: a working day cannot cross midnight, so night hours need two slots; and the clones lost a setting and started switched off. One setting showed as on, but we could not find where to change it.

• Human score: 5 / 10 (25 of 50: night hours 3, making meetings 8, clones 4, finding a setting 2, look and speed 8).
• AI-friendly score: 5 / 10. Good: buttons have names, and there is a public API. Bad: the page text came back empty on load, and some sections needed two tries to open.

SciSpace, API access form (05.11.2026)

• Human score: pending (we're waiting for their reply).
• AI-friendly score: 6 / 10. Good: clear labels, and the form opens at its own address. Bad: on the SciSpace page it sits in a frame AI tools cannot see, and the tool picker allows one choice when we needed several.