Gaps and limitations, observed from the repository. Concrete defects are in
internal/known-issues.md.
Build files carry a current price with no indication of when it was fetched. If the daily workflow has been failing for a month, the numbers look exactly as authoritative as fresh ones.
A "prices as of" line per file, written by the scraper, would fix that in a single change and is the most valuable item here.
Because the scraper rewrites prices in place every day, the git history already contains a daily price series for every tracked component going back to when automation started.
That is a genuinely interesting dataset — component depreciation over years — and nothing surfaces it. Extracting it would need no new collection, only reading what is already committed.
The build index is maintained by hand. It lists every build with links to three
representations, and nothing checks that the Markdown links resolve or that every file appears.
It has already drifted once — see internal/known-issues.md.
One upstream, no fallback. Every current price comes from PCPartPicker. A layout change there stops all price updating, and there is no secondary source.
Discontinued parts degrade quietly. An unavailable component simply shows no current price, which is indistinguishable from a scrape failure.
No coverage record. Nothing states which issues are archived and which are missing, so a gap in the years looks the same as an issue that had no build.
The website build is not automated alongside prices. Prices update daily on main;
gh-pages can lag.
- Being a shopping list. Most of this hardware is discontinued. The archive exists for the comparison between printed and current prices.
- Reproducing the magazine. It records which parts each build specified, not the articles around them.