What can be done locally?
Explore the boundary between private, continuously available local inference and elastic cloud capability.
Starting point
Local models may be best for private, continuous work with predictable marginal cost, while cloud models can provide burst capacity or capabilities the local cluster cannot match.
What to track
- How powerful an air-gapped software factory can become.
- The best use of slow but continuously available local inference.
- Whether a near-frontier local model should drive the factory or optimize dispatch.
- How the boundary moves with hardware, model quality, latency, and API prices.
Living record
Local-model evaluations, hardware changes, and air-gapped experiments belong here.