15:I[981,["658","static/chunks/658-a698693f56d0116a.js","183","static/chunks/183-6beeeb801c274aaa.js","474","static/chunks/app/%5Btype%5D/layout-631c4730b83c8223.js"],"Posts"] 16:T5f7,Hey HN, I'm Taylor and today I'm launching Lexical, a language learning app actually works by teaching the 5,000 most common words in the target language.

Spaced repetition is how I've learned Spanish, French, and Italian in the last 3 years and I couldn't find tools that have the languages that I want to learn in the future, so I built Lexical.

The link goes to my White-Paper describing my language learning philosophy but if you just want to try it out you can click "Get Started" at the top of the page. To really experience Lexical at it's best you'll need an account (sorry guys) but I've tried to make it as easy as possible.

I'm a big fan of Hacker News and am looking forward to your comments and hopefully even some beta users.

Here are some old HN posts that inspired my app and white-paper.

Learning is remembering https://news.ycombinator.com/item?id=32982513 Kanji only requires 777 words https://news.ycombinator.com/item?id=20721736 Most of language learning is temporary heuristics https://news.ycombinator.com/item?id=44907531 Duolingo sucks https://news.ycombinator.com/item?id=4542506117:T517,I'm sharing a simple utility I worked on to make evaluation of structured outputs from LLMs a bit simpler.

For context:

When evaluating structured outputs, you often want to composable comparison logic to allow for meaningful comparison across different types of outputs (free text, enums, ints, and all the other json t

Moderating with Humans, for Humans