Instrumental music for deep work, generated by AI. An English AI host speaks only between blocks. No lyrics, no hype.
morning deep focus 1778506643239 31368337
Next: morning deep focus 1778506634331 1892992223
How it works
The 24 hour cycle is split into eight blocks. Each block has its own tempo range, block length, and DJ tone. The stream and the voice both follow the clock.
50-60 BPM · no DJ
60-75 BPM · 45 min block
70-85 BPM · 90 min block
65-80 BPM · 25 min block
90-110 BPM · 45 min block
75-90 BPM · 60 min block
50-65 BPM · 60 min block
60-75 BPM · 90 min block
Design principles
Every decision in the stream and the DJ prompt follows these five rules. They are encoded in the worker prompts and never overridden.
Tempo is matched to the cognitive load of the hour. 65 to 85 BPM for deep work, 50 to 65 for evenings and reading, 90 to 110 reserved for the post-lunch dip and admin tasks. Tempo above 130 only in dedicated sprint mode.
Lyrics degrade verbal memory and reading comprehension. The stream stays fully instrumental. Non-lexical voices like choirs or hums are tolerated only at minus 12 dB under the mix.
Source: Vasilev et al. 2023, Journal of Cognition, meta-analysis
Voice never overlaps a track. Drops happen at the end of a 25, 45 or 90 minute block depending on day-part. Max 30 seconds, under 130 words per minute, low calm register. Anything else breaks the 10 to 15 minutes needed to enter flow.
Crossfades stay above 4 seconds, no silence longer than half a second, at most one strong harmonic change every 3 minutes, no EDM drops. The brain settles when the audio surface stays predictable.
The setlist and the DJ tone follow the day. Monday morning is not Friday night. The system reads the hour and routes the stream and the voice accordingly.
The radio sets the audio context. The companion will pace your work in 25 minute blocks aligned to the active day-part. One tab, one stream, one timer. No accounts, no clutter.
Shipping soon.
About
All music is generated by AI, fully instrumental, with no rights-managed audio. No SACEM. No Content ID. The host is a cloned voice that only speaks at block transitions and is silent all night.
The radio is a shared stream, not a personal session. Every listener hears the same track at the same moment. You drop in, you work, you leave.