shitfa.stRoast Review

The hero announces An archive of landing pages that all read like one AI wrote them, which is a remarkably strong confession to place above the fold when the rest of the page has one unbroken editorial voice. The copy is sharp, the labels are real metric[redacted] consistent, the jokes land on schedule, and even the breaking-news ticker has a repeatable format for reporting purple-gradient sightings. It does not feel like a team of writers made a site; it feels like one very confident prompt discovered a content calendar.

The premise could have been a gallery of bad homepages, but shitfa.st adds Featured Offender, Freshly Accused, Taxonomy, Score It Yourself, Weekly Awards, Elite Slop, evidence logs, quick filters, score dimensions, marginalia, case files, and a field guide to post-boilerplate sameness. The site takes a simple joke and gives it a governance layer, complete with categories for the categories and an editorial desk for the evidence of the editorial desk. Somewhere, an LLM is being told to take this concept further and has just invented a committee.

The submitter is also correct about the machinery: there is an AI roasting procedure and ranking system behind the scenes, but it was entirely created and is guided by the editors of this website — us. That makes the archive less an untouched machine judgment than a curated human-plus-AI editorial format, which is more honest and somehow more incriminating: the uniformity, density of punchlines, and suspiciously complete scorecards are not accidental residue. We built the apparatus, taught it the house voice, and then pointed it at every other apparatus.

The footer says We get irony., and to its credit, it does. The line is the site equivalent of a restaurant putting a framed health-inspection certificate next to a grease fire: technically transparent, visually composed, and not evidence that the kitchen has stopped smoking. The irony is the product proof, because shitfa.st really does have screenshots, real entries, and useful pattern language; it just also proves that an archive about AI sameness can become a highly organized monument to its own particular brand of sameness.