HackerBooks: what a synthetic UX audit uncovered
I'm in my second year studying CS and I want to read beyond my textbooks. I spend most of my free time on Hacker News and Reddit learning about what professionals actually read. I have a limited budget so I need to be selective about which books I actually buy.
HackerBooks is a web product that drew attention this week: Launch. We ran a synthetic UX audit on it, sending a persona suited to the site through its pages and recording what it hit.
What our persona hit
- product-grid section. Each card's only evaluative signal is a tiny orange '94 recommend · 172' line, which has no label explaining what the two numbers mean. The first is probably a recommend score and the second a mention count, but the page never says so. There is no scale, no star rating, no difficulty or level tag, and no one-line description. A buyer on a budget must guess what 'recommend' measures and whether a book suits their level, so they cannot compare 'Effective Java' with 'Pro Git' without leaving the page. Heuristic: NN/G Heuristic #6, Recognition rather than recall.
- product-grid section. Each card carries only a cover, title, author, and a line like '237 recommend · 339'. The grid mixes Hamlet, Siddhartha, Starting FORTH, Hacker's Delight, and The AWK programming language with no difficulty, format, or one-line description. A buyer on a budget cannot judge level or fit without leaving the grid. The O'Reilly-style skill tags and star ratings this audience expects from Goodreads and Amazon are missing. Heuristic: NN/G Heuristic #6, Recognition Rather Than Recall.
- product-grid section. The social-proof line pairs two unlabeled numbers, '237 recommend · 339'. The second number has no label or tooltip, so it could be mentions, comments, or a total. The orange recommend count reads like a rating, but there is no scale, so it gives no basis for comparing a 15-recommend book with a 237-recommend one. Heuristic: NN/G Heuristic #2, Match Between System and Real World.
- product-grid section. The page presents two pill filter rows with 10 and 4 options, followed by a sidebar of 20+ topics with counts in the thousands. The pill rows are clean and clearly labelled 'PUBLISHED' and 'FIRST MENTIONED', so that part is easy to understand. The sidebar topic list, however, runs past the fold and has no grouping, so the user has to scan a long column. Counts like '3,273' also add noise to every row. Heuristic: Hick's Law.
- product-grid section. The covers form a clean, evenly spaced nine-column grid. Each card repeats the same year-range badge, title, author, and stats in the same position, so scanning is easy. However, the covers are loud and wildly varied, and the small gray author and stats text recedes beside them. The 'Gulag Archipelago' placeholder cover is a plain maroon tile and looks broken next to the real covers. Heuristic: Gestalt, Similarity.
This is a single-page heuristic pass by a synthetic persona, not user research; it can miss context and occasionally misreads elements.
This audit was run independently by Blinx; HackerBooks was not involved.
See a complete run end to end: a full example Blinx report.
Curious what a persona suited to your site would hit?
Run this on your own site