The catalog used to greet you with the most expensive machines on top — a $5,000 workstation is not what most people are shopping for. It now leads with a Score: one number for how good a pick a mini PC actually is.
Score rewards what makes a mini PC worth buying — raw CPU performance (PassMark), a fair price, a small footprint, and the RAM & storage you actually get — and ignores brand and seller opinion entirely. Higher is better. Sort any column to slice it your way, and hover the number to see exactly where the points came from. The full formula lives on the About page.
A couple of smaller things that landed recently too: a “sign in to fix this spec” prompt on model pages (spot a wrong number, correct it on the spot), and a pending-confirmation banner when you change your account email.
Under the hood
The Score is computed in your browser over the whole catalog and stored nowhere, so it can’t go stale — and it’s absolute, not a percentile ranking. That’s deliberate: a relative score drifts as the catalog grows (a model could read 64 one day and 55 the next without anything actually changing), whereas an absolute one only moves when that machine’s own price or specs change. Performance is linear in PassMark, so the scale is uncapped and climbs as CPUs get faster; price and size use diminishing-returns curves; RAM and storage saturate (a barebone with no RAM/SSD simply earns nothing there). Energy efficiency joins the formula once we have power figures for enough models.
Also quietly buried this round: a view-counting experiment. It turns out a Directus flow can’t bump a counter without writing an audit-log row for every single page view — which would balloon the history table forever — so for a fairly weak signal it wasn’t worth it. Deleted, maybe revisited properly later.