I went quiet for a while. This is what I was building.
I have been quiet here for a while. This is what I was doing. theknowngood.com is live. It tracks published evaluation data for AI models: benchmark results, pricing, performance, arena ratings. Around 900 models, roughly 2,300 static pages. CSV and JSON exports are free and ungated under CC BY 4.0. No accounts, no tracking, no API. I am a systems engineer rather than a developer, so the…
For quite some time, I have kept quiet about a significant project I was working on. That project is now live at theknowngood.com. This website tracks evaluation data for AI models, including benchmark results, pricing, performance, and arena ratings. It features around 900 models, with approximately 2,300 static pages. Users can freely download CSV and JSON exports, all under the CC BY 4.0 license, without needing any accounts, tracking, or API access.
As a systems engineer rather than a developer, I focused on architectural decisions. The website's data source is PostgreSQL, but it does not interact with the database during request times. Python and Jinja2 are used to render the entire site into flat static HTML files on a schedule. This approach ensures repeatability, as the static build is reproducible and diffable, and it is also cost-effective, as flat files remain affordable even under heavy load.
My hosting strategy involves no inbound ports; nginx binds to loopback only, and the site is accessed through an outbound tunnel. This setup eliminates a category of potential issues, opting instead to defend against problems rather than prevent them. Every chart on the site displays its own denominator, showing Top N of M, with the actual M, not the total population. This detail was the most crucial correction I made throughout the project.
Please note that this website is self-hosted on hardware in my home. If you find it slow, feel free to inform me so I can investigate. I prefer receiving feedback directly from you rather than discovering an issue in logs three weeks later. Your feedback is highly appreciated, especially regarding tests I may have overlooked. A newsletter option is available on the homepage, also accessible without any account.
Additionally, I documented every challenge faced during the development process, including instances where most of the implemented checks passed while measuring nothing. This information is available in the link provided in the reply below.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
