Four Tools in Seven Days

Every week I build a stack of things with an AI agent, and most of it dies in a closed terminal tab. This is the receipt: what got made this week, and why it was worth making.

Four projects stood out.

What got built this week

An NBA "Data Lab," built to cheer up my mom

The Cavs got swept and my mom was crushed — Donovan Mitchell is her favorite. So I built a data page arguing he's a top-10 player, hard numbers only.

The interesting part was the metric. A plain points-and-assists ranking doesn't get him there, so I iterated until I had an availability-weighted score: production discounted by how reliably you're healthy enough to deliver it. Durability is value — and it happens to make the case for Mitchell.

That page became a reusable NBA Data Lab: a top-20 ladder, a "Health Curse" visualization, and a playoff-injury study back to 2010. Live, and shared with the family.

A family travel map spanning 35 years

The week's busiest project. I added my grandmother's 1987 Europe backpacking loop, six family trips from 2005–2018, and a sibling's 42-stop run across Asia — decades of family movement on one screen.

New features: a distance-covered calc for road and backpacking routes, and a per-person stats view with a selector. The photos ship through a private pipeline — public repo, private images — so I can share the map without putting family pictures on the open web.

A dive map that became a 4K archive

I loaded 117 4K clips onto the dive sites and curated a best-of: mantas, turtles, a wall of fish, dolphins.

Then I built the analytics in two layers — standard Google Analytics, plus a custom Cloudflare worker that flags returning visitors by IP. So I can see who comes back to the map, not just who shows up once.

A YouTube tool that tells me why a video flopped

A pipeline for Twitter, Threads, and YouTube — but the useful part was a channel-assessment tool that scores each video and cleans up its first and last frames (the thumbnail and the sign-off). I ran it across 25 clips.

When a fresh upload tanked, I pointed the tool at it and got an answer instead of a guess.

By the numbers

Commits per day

Quiet stretch, then a Sunday–Monday surge when the Lab and the dive archive both landed.

Where the commits went

This newsletter was auto-assembled from my own session logs — so it's also one of the things I built this week.

Next week: more Lab pages. — rickleberry