One language, deliberate execution choices

Aner currently parses, checks, and interprets source through a C++ executable. Compiling that interpreter is distinct from compiling an Aner program. Native compilation of user programs is a later milestone.

The machine learning design proposes a shared tensor interface with explicit device placement and backend contracts. Apple GPU, NVIDIA GPU, lower precision, multicore scheduling, and distributed execution are research and engineering work ahead.

A credible evaluation must keep correctness, conversion behavior, reproducibility, and workload boundaries visible alongside performance. There are no published speedup, scaling, energy, or hardware parity results to report.