Longer-form thinking from fraQtl — inference economics, task-aware compression, mathematics, and whatever else the work surfaces. The receipts live on the benchmarks page; this is the reasoning around them.
The first race was to create more intelligence. The second is to deploy only as much of it as each task requires. An offline iPhone demonstration, the economics of inference at adoption scale, the physical supply chain behind every token — and why the winning metric is cost per successful task, not cost per token.
More notes in progress · contact@fraqtl.ai