Functional music, honestly designed.

CognitionFM renders original, long-form audio for focus, stress downshift, and sleep. Every design decision traces to published research, and every mix is reproducible from a seed.

Read the evidence See the playlist designs

No 432 Hz. No solfeggio. No binaural-beat promises. No "dopamine hacks." Where the evidence is weak or mixed, we say so directly. For complex verbal work, silence might beat any music, including ours: test it.

The formats

Deep Work - Verbal

Writing, reading, hard debugging. No lyrics, no surprises.

Deep Work - Analytical

Routine coding and flow work. One steady, soft pulse.

Downshift

25 minutes of descending energy for stress downregulation.

Sleep Wind-Down

Pre-bed routine audio that fades to near-silence.

Morning Ramp-Up

A gentle arousal ramp into the day.

What the design rests on

Major claims are cited to peer-reviewed research, prioritizing systematic reviews and meta-analyses. The five sources doing the most work:

Kämpfe, Sedlmeier & Renkewitz (2011), meta-analysis, Psychology of Music: background music has small, task-dependent effects; it hurts reading and memory, helps mood and sport.
de Witte et al. (2020), two meta-analyses, Health Psychology Review: music interventions reliably reduce physiological and psychological stress.
Jespersen et al. (2022), Cochrane systematic review: listening to music improves subjective sleep quality in insomnia (moderate certainty).
Terry, Karageorghis et al. (2020), meta-analysis of 139 studies, Psychological Bulletin: music measurably improves exercise performance and reduces perceived exertion.
Salamé & Baddeley (1982) and the changing-state literature: unpredictable sound, including lyrics, disrupts verbal short-term memory even when ignored.

The full claim-by-claim review, including what we consider weak (binaural beats, creativity effects, intimacy) and debunked ("healing frequencies", the Mozart effect as popularly claimed), is on the evidence page.

Why trust this

Evidence, tiered. Every claim in the review is labeled well-supported, moderate/mixed, weak, or debunked, with sources you can click.
Ownership, provable. All audio is synthesized by our engine. The ledger lists recipe, seed, and engine version for every published minute; identical inputs regenerate the identical mix.
Tested, not asserted. Formats ship only after beating or tying a commercial reference playlist in our own n-of-1 protocol.

Status

The engine, recipes, and documentation are complete and current. First public releases are in preparation; this page will link to them when they exist, and not before.