Meet Murphy.

No — and that's the wrong target. The realistic and valuable goal is autonomous-enough that the engineer focuses on the 20% that actually requires their ears, while Murphy handles the structural decisions reliably and flags anomalies. Murphy will never be your favorite engineer. He'll be a co-pilot who handles gain staging, feedback, and level riding so you can focus on the artistic calls.

Yes. The bar isn't "replace a pro engineer" — it's "be better than a vocalist running back to the board between songs." That bar is low and Murphy clears it. Hold the gain structure from soundcheck, suppress feedback in 250 ms, ride the vocal when the singer steps back from the mic — all of this works right now without a measurement microphone.

Feedback suppression. No close second. The vocalist can't stop singing mid-note to find and notch a frequency. The drummer can't reach the board. Murphy catches feedback in 250 ms and applies a narrow EQ notch before the audience hears the squeal. This single behavior justifies the product for any band running their own mix.

Through engineered audio features, not raw waveforms. A real-time analysis pipeline turns audio into numerical descriptors that capture what human ears perceive — spectral content per channel, masking matrices between instruments, dynamics, perceived loudness (LUFS), and room transfer functions. Murphy reasons on those features the same way a human engineer perceives them, just more quantitatively. The quality of the feature extraction is the quality of his hearing.

Eventually, for room correction at pro venues. Not yet for small-band self-mix. A measurement mic captures what the audience actually hears — useful for room correction, house feedback detection, and the perceptual ground truth. But most small bars don't have a reference for that anyway, so Murphy operating without one is still a massive upgrade over what the band has today. The room mic is a tier-2 capability for bands that grow into it.

For your first show in a brand-new room with no training data, Murphy is a sophisticated safety net — gain staging, clipping protection, feedback suppression, level riding. For show 50 with the same band in a venue Murphy has seen before, Murphy is a co-pilot that handles 70% of the decisions you would have made anyway. The asymptote isn't "fully autonomous in any room." It's "autonomous enough that you focus on the 20% that actually requires your ears."

The system is designed around three principles: conservative, observable, recoverable. Corrections are small and trend toward the soundcheck baseline — never large sudden moves. The band sees in real time what Murphy is doing and why. And there's a one-tap "Reset to Soundcheck" button on the iPad — labeled for the guitarist between songs, not a tech. If anything ever feels off, one tap returns the board to where you set it during soundcheck.

Artistic decisions. "Make the guitar warmer." "Compress the snare with character." "Build the chorus differently than the verse." These require aesthetic taste that doesn't reduce to a feature vector. The perceptual ground truth for those calls is subjective and context-dependent in ways autonomous mixing won't ever fully capture. Murphy is deliberately built to leave those decisions to the engineer.

Human-touch lockout. When a band member nudges a fader or a send, Murphy freezes that channel and yields to the human until the band confirms. The lockout is visible in the UI — "Ch 5: Human hold (14s remaining)" — so the band can see Murphy is respecting their move. The system is built so Murphy never overrides a human decision silently.

Two modules fighting each other in a feedback loop. Murphy's vocal-level rider boosts a quiet channel. The boost trips the feedback detector. The suppressor applies a notch. The rider perceives the level drop from the notch and tries to compensate. The detector fires again. This can get loud and stay loud while Murphy fights itself. The fix is a hard interlock: when the suppressor is active on a channel, the rider on that channel freezes. No exceptions. These two modules know each other's state at all times. This is the single most important safety interlock in the system.

Three reference points: per-channel pre-DSP signals (the raw inputs from each instrument), the post-mix bus (what's being sent to the PA), and — for the room-correction layer — a measurement microphone at front-of-house position. The intelligence lives in the deltas between these three: channel-to-mix tells Murphy what mix decisions exist, mix-to-room tells him what the PA and venue are doing to those decisions, and channel-to-room tells him the full physical reality.

Both, but the value calculus is different. For pro touring engineers, Murphy is a co-pilot that handles the 70% of structural decisions so the engineer can focus on artistic moves. For small bands self-mixing from the stage, Murphy fills a gap that nobody else fills — the cumbersome adjustments the band physically can't do while playing. The small-band scenario is realistic right now with current sensors. The pro tour scenario asymptotes to co-pilot, not autopilot, and that's the right product shape.