3 ms·
If only we had existing formats and standards, that LLMs were probably already trained on, which allowed for serializing components and their layouts.
by ActionHank 18d ago
If only we had existing formats and standards, that LLMs were probably already trained on, which allowed for serializing components and their layouts.
- vkrychun 18d agoWell, StemJSON is also built on an existing format - JSON, which LLMs are definitely trained on and generate quite well. The key is the vocabulary, which describes not only components and layout but also their behavior, so in the end you get pretty comprehensive functionality or even a complete app.