lemon

Huh… TIL other countries do Sinterklaas as well. Miklavž… That’s Slovenian, going by the wikipedia page?

Here in Flanders and to the north in the Netherlands we actually get told this story. Sinterklaas, who visits on the eve of the 5th (NL) or during the night (BE) is our OG Santa Claus.

For kids Sinterklaas is the main event, whereas Christmas is when you get given… socks. Last few years the focus seems to have shifted more to Christmas though.

Glad we got rid of ‘Zwartepiet’ though (Sinterklaas’s helpers). Think ‘elves’ except human-size, not-elf and with blackface.

I grew up attending international schools. Seeing my Dutch classmate dressed as zwartepiet getting called to the principal’s office is a core memory.

SPA websites that go out of their way to make opening a link in a new tab difficult to impossible can well and truly go fuck themselves

ML engineer here. My intuition says you won’t get better accuracy than with sentence template matching, provided your matching rules are free of contradictions. Of course, the downside is you need to remember (and teach others) the precise phrasing to trigger a certain intent. Refining your matching rules is probably a good task for a coding agent.

Back in the pre-LLM days, we used simpler statistical models for intent classification. These were way smaller and could easily run on CPU. Check out random forests or SVMs that take bags of words as input. You need enough examples though to train them on.

With an LLM you can reframe the problem as getting the model to generate the right ‘tool’ call. Most intents are a form of relation extraction: there’s an ‘action’ (verb) and one or more participants (subject, object, etc.). You could imagine a single tool definition (call it ‘SpeakerIntent’) that outputs the intent type (from an enum) as well as the arguments involved. Then you can link that to the final intent with some post-processing. There’s a 100M version of gemma3 that’s apparently not bad at tool calling.