Stories by stephantul

From Chesterton's fence to Chesterton's gap

Why scikit learn's fit transform is probably not for you

Show HN: Semble – Code search for agents that uses 98% fewer tokens than grep

Show HN: Semble – Fast code search for agents with near-transformer accuracy

Show HN: Skeletoken, a Python package for editing model tokenizers

Show HN: PyNIFE. 400-900× speedup for embedding-based retrieval pipelines

Show HN: Skeletoken, a Package for Editing Tokenizers

Turning any tokenizer into a greedy one

Decasing Transformers for Fun

Model2Vec as a Fasttext Alternative

Using overloads to handle union return types in Python

Ask HN: Favourite resources for learning programming type theory?

Evaluating ML classifiers using relative error instead of absolute accuracy

Defeat stringly typing without making your users unhappy

Distilling ModernBERT into a static model doesn't work

Show HN: SemHash – Fast Semantic Text Deduplication for Cleaner Datasets

Train faster static embedding models with sentence transformers

Semhash: Fast deduplication and dataset multitool in Python

Model2Vec: Make sentence transformers 500x faster on CPU, 15x smaller

Show HN: Model2Vec: make sentence transformers 500x faster on CPU, 15x smaller