How to run useful AI on hardware you already own, or can buy second-hand for the price of a year’s subscriptions. Every number in these was measured on my own machines on the day of writing — not quoted from somebody’s benchmark.
One PHP file, one database table, a spam check and an email provider. It stores every enquiry before it tries to email you, so nothing gets lost. It refuses anything it can't verify, and it writes down every refusal with the reason. The one on this site has turned away 164 robots since August, and I can tell you why it refused each of them.
Seven pieces, twenty-one and a half minutes, one house theme, written in an afternoon for my game Rift Haul. It cost nothing and needed no music generator and no graphics card. Every note was written as data by an AI that cannot hear, and performed by free sampled instruments on a Raspberry Pi. Here is how it was done, and a script that renders the opening of the theme on your own machine in about fifteen seconds.
A folder of Markdown chapters goes in, and a validated EPUB and a print-ready 6×9 PDF come out. You need no pandoc, no Calibre and no LaTeX, and you pay nothing. Two short Python scripts, the official validator, and the browser you already have. Both of my books are built by a bigger version of exactly this, and here is the small version, built from scratch and run this evening.
Jellyfin on a Pi 5 is up in three seconds, costs 238MB of memory and scans a library in seconds. It is also 100 times slower at transcoding than at not transcoding, and the settings page will cheerfully offer you hardware acceleration this machine cannot do. Here is the whole build, the one dropdown to leave alone, and the query that found a library of 19,935 items of which three existed.
Two database tables, one shell hook and about fifty lines of PHP will tell you what every AI agent you run has been doing. Mine has recorded 57,875 heartbeats since April for 76MB. Here is the whole build, command by command — then the six ways it lied to me, every one of which reported success, and the six queries that catch them.