Compose first, respond second
The default After Hours scene is a composed 32-bar record built from felt keys, muted bass, brushes, and a recurring hook. Typing pace, pauses, and agent outcomes influence its arrangement at musical boundaries. Individual characters do not choose its notes.
The original character instruments remain as Classic styles, where keys propose tuned, quantized melody notes. Keeping those modes distinct makes the interaction understandable: one shapes an arrangement, while the other plays an instrument.
One engine, two ways to listen
The browser version is a page to type into, with audio and text processing in the browser. The local session version uses a Node hub to connect system-wide typing activity and coding-agent events to the same Web Audio engine.
In After Hours, errors and successful outcomes update aggregate tension and resolution. The musical director responds at the next appropriate boundary. That preserves a composed sense of time even when the work itself is irregular.
Save the performance as a score
The session shelf records resolved sound events rather than typed text or keys. A saved tape can be played again through the engine or rendered to a WAV file with an OfflineAudioContext.
That is a re-performance rather than a recording of the original audio waveform. Note events stay fixed while some tape character can vary. It is a useful separation: the record of the session contains the musical decisions, and the synthesis engine turns those decisions into sound.