GitHub - NandhaKishorM/laya at console.dev

kept by eddie oct 02

Laya provides multilingual, non-autoregressive typed decision making (choice, score, yes/no) in a single forward pass, routing to language‑specific checkpoints for ~33 ms latency.

tokens too cheap to meter
The price of running machine‑learning intelligence is falling by orders of magnitude yearly, making AI tokens as cheap as electricity and enabling pervasive local and hosted AI use.
also on Laya, #open-source, #developer-tools
GitHub - cased/kit: The toolkit for AI devtools context engineering. Build with codebase mapping, symbol extraction, and many kinds of code search.
Kit provides a Python toolkit for codebase mapping, symbol extraction, and code search to build LLM-powered developer tools and agents.
also on PyPI, #open-source, #developer-tools
GitHub Actions is the weakest link
GitHub Actions' default features—mutable tags, pull_request_target, template injection, and write-scoped tokens—have enabled a chain of supply-chain attacks on open-source registries, and GitHub's opt-in security roadmap won't fix the root causes.
also on PyPI, #open-source
How Jev works: calibrated decision models
Jev is a low-latency decision model that scores options via LLM forward passes to output calibrated probabilities without text generation.
also on Hugging Face, #machine-learning, #ai-integration, #developer-tools
ZeroModels
ZeroModels provides 100+ pretrained transformer models in pure Keras 3 that run on JAX, PyTorch, or TensorFlow without requiring transformers or torch at runtime.
also on Hugging Face, #machine-learning, #open-source, #developer-tools