This has always irked me too. Closest thing I can think of are the complex behaviours of some insects (like Sphex wasp en.wikipedia.org/wiki/Sphex) that are fine-tuned to the environment, but don't result in permanent behavioural changes. D. Dennett called this mindless intelligence "Sphexishness".

Read the cookbook: arxiv.org/abs/2510.00368

Join us for weekly seminars on formal language theory, ML, NLP, and more: flannseminars.github.io

There is no better way to understand what transformers can do than to get your hands dirty and construct them, weight-by-weight. The Transformer Cookbook provides a guide for anyone aiming to understand the expressive power of transformers on such a formal level.

Thanks to all the chefs: @ccwatson.bsky.social, @antonxue.bsky.social, @satwik77.bsky.social, @ll4r3n4.bsky.social, @lambdaviking.bsky.social, Emile Dos Santos Ferreira, @anejsvete.bsky.social, @dchiang.bsky.social

Neural methods haven't taken over as much you'd think, especially in smaller companies that can't afford large-scale data + compute. OpenFST is still used directly in Hybrid DNN-HMM ASR models, and also indirectly, through Pynini, in (Inverse) Text Normalisation, needed for both TTS and ASR

You might find these useful mirror.aclweb.org/emnlp2014/tu..., borjaballe.github.io/slides/fopss...