[–] librasteve 4mo ago ↗ DSLs are a secret weapon for LLM effectiveness because their human-readable, domain-centric structure constrains both the training set and model outputs, making them significantly easier for LLMs to generate accurately.
[–] apogee 4mo ago ↗ Raku grammars are cracked. Here's Jinja2: https://github.com/m-doughty/Template-Jinja2/blob/master/lib...
2 comments
[ 4.2 ms ] story [ 18.8 ms ] thread