Dataset and evaluation
Data health, split integrity, baselines, holdouts, task metrics, regression gates, and variance.
Tensor Cortex Insights
Field notes for building better open models on your Mac.
Evidence-led technical guides for developers, researchers, and small AI teams working with local models. Each article defines the decision, cites primary sources where they matter, and contributes a practical framework, protocol, or reproducible artifact.
Scope
Four connected parts of one trustworthy model workflow.
Apple Silicon and MLX
Memory, throughput, quantization, hardware-aware planning, local training, and Multi-Mac direction.
Fine-tuning and distillation
Choosing the right method, designing bounded experiments, and comparing teacher, baseline, and candidate models.
Packaging and reproducibility
Project manifests, checkpoints, model cards, provenance, portable formats, and known limitations.
Articles
Newest research
Why You Need a Baseline Before Fine-Tuning a Local LLM
A practical protocol for measuring the base model, protecting the holdout, and deciding whether a local fine-tuning experiment actually improved the task.