Colophon
Scoring
We read every page in a company’s sitemaps whose address looks like a comparison, keep the ones that compare it with a named rival, and ask five yes/no questions of each. Every yes is a red flag.
- Wins on everything Does the page claim the vendor wins or ties on everything it compares? 1
- Never says when to pick the rival Does the page leave out every situation where a competitor would be the better choice? 2
- Admits no weakness Does the page avoid saying anything the vendor's own product can't do or does poorly? 3
- Vague attacks Are the criticisms of competitors vague, with no specifics, numbers or sources? 4
- Rival gets more criticism Does the competitor get clearly more criticism than the vendor's product? 5
A model answers each question, and an answer counts as a red flag when it’s at least 80% sure. Companies are ranked once 3 pages are read; one red flag per page or fewer is fair. Before each version goes live, the scoring is tested against pages labelled by Claude and cross-checked by two other models.
Type
Both are Whimsical’s open-source fonts, under the SIL Open Font License. This site serves them itself.
Colour
One colour per question, in the same order everywhere: dots, bars, legends and the icon.
- Question 1 Wins on everything
#E9631C - Question 2 Never says when to pick the rival
#2D63E0 - Question 3 Admits no weakness
#12A06A - Question 4 Vague attacks
#3B1E6E - Question 5 Rival gets more criticism
#EC3E7B - Ink Text, rules, buttons
#1A1630 - Muted Secondary text
#5F5A72 - Line Row separators, borders
#E6E3EC - Panel Boxes
#F4F2F8 - Fair Fair players
#1C6B4A
Icon
Five questions, five dots on a circle, 72° apart: orange at the top, then clockwise in question order.
Built with
- Site
- Astro, built to static pages. Sass for styles, a few lines of TypeScript for sorting and filtering.
- Data
- One SQLite database, read when the site is built.
- Share images
- satori lays each card out, resvg turns it into a PNG, at build time.
- Crawler and pipeline
- Python with uv, httpx and Beautiful Soup. It runs every Sunday on GitHub Actions.
- Scoring
- OpenAI’s gpt-6-luna, methodology v8.
- Checks
- Biome and astro check for the site, Ruff for the pipeline.
- Hosting
- Cloudflare Pages.
Data
Every week’s results are published as a SQLite database and CSV files. Every company’s page is also
available as JSON, at its address plus .json.
Privacy
The site sets no cookies. It counts visits with Plausible, which uses no cookies and collects no personal data. The fonts and images are served from this site.
Credits
Made by Whimsical.
Code, method, benchmark labels and data under the MIT license. Fonts under the SIL Open Font License.