Source, protocols and measurement records for comparing Lloyal's execution model with conventional agent applications.
| Experiment | Status | Start here |
|---|---|---|
| Memory | CUDA capacity measurements, Fieldnote workload comparison and runnable benchmark setup | Setup and commands |
| Reasoning | Historical exploratory observations; dedicated causal experiment pending | Existing evidence and open controls |
Each experiment owns its dependencies, fixtures, protocols and results. Run commands from that experiment's directory. Recorded results retain their original scope and limitations; setup portability changes do not constitute a new benchmark run.
The memory experiment includes the pinned Fieldnote application and its web UI, an idiomatic LangGraph JS host, and native/server instrumentation. Large runtime artifacts, model weights and raw session archives are excluded from Git. The records identify what is included and what still requires a separate evidence archive.