Building the Starlog database
This site already looks like a starship console, so the only honest direction left was escalation. I love Star Trek, and it bothered me that decades of captain's logs, the franchise's signature storytelling device, existed nowhere as a thing you could actually search. So I built it: every log entry from the classic era, TOS through Enterprise plus the films, in one database at /starlog. Nobody asked for this. That has never once stopped me.
The build ran on the same playbook as the job hunt from the last entry: a fleet of AI agents doing the typing while I made the calls and signed off on the data. The raw material was fan episode transcripts, seven hundred-odd episodes' worth, and the extraction is pattern matching that sounds trivial until you meet the data. Curly apostrophes were silently swallowing entries because a right single quotation mark is not, to a regex, an apostrophe. The films defaulted every captain's log to Kirk, which quietly misfiled five of Picard's, and I am not sure either man would have been pleased.
My favourite bug survived every automated review and fell over in the final walkthrough. The speaker filter showed sixteen names, taken alphabetically, and the alphabet being what it is, the list included Aquiel, a character with one episode to her name, but not Picard, who at three hundred and ninety log entries is the most prolific logger in the classic era. A database of captain's logs in which you could not filter for the captain who made the most logs. It filters by frequency now, as it should have from the start.
Somewhere in the quality pass, checking entries by hand, I found Odo's first ever security log, recorded at Sisko's request and visibly under protest: humans have a compulsion to keep records and lists and files, so many that they have to invent new ways to store them microscopically. I read that sentence while sitting on top of a JSON file containing one thousand one hundred and eighty-eight of Starfleet's records, compiled voluntarily, for fun. The man has a point and I have chosen not to dwell on it.
It's live now: full-text search, filters by series and speaker and log type, stardate lookup, and a random entry button, which was the cheapest feature to build and is obviously the best one. The streaming-era shows aren't in there yet, mostly because reliable transcripts aren't, and whether that itch gets scratched is a question for a future entry. In the meantime, if you've ever half-remembered a log entry and wanted to find it, the computer will now take your query.