In this rig frequency-budgeting and masking-avoidance are predictive — no framework surfaces a wired metering or perception bridge

The source flags a rig limitation: there is no true master bus, EQ, or spectrum analyzer surfaced to the agent yet, so mix reasoning is predictive — arrange to avoid masking, and use per-voice filters, gain, pan, and sidechain-shaped ducking. Even SuperCollider and ChucK, which have FFT UGens, never route them to a perception layer. The source therefore marks frequency-budgeting and masking-avoidance as approximate in all five audio realizers — not because the frameworks can’t EQ or filter, but because no wired perception bridge exists. This is blocked on the L3 perception layer, not on any DSL capability; until L3 exists, all mix decisions are intent-based, not measured.

Examples

A Strudel agent writing .hpf(300) on a pad is making a predictive arrangement decision (intent: remove mud-zone energy), not a measured EQ move — there is no spectrum readout to confirm the result.

Assessment

Why is frequency-budgeting marked ‘approximate’ even in SuperCollider, which has FFT analysis? What is the L3 perception layer and why is its absence the root cause?