Picking the right content
stops depending on who is on shift.

01Where it lives

Both surfaces use the same engine, inside the device. They cannot give different answers.

02How it decides

1. 01
   What the team wrote about the content

   Category, intent, clinical tags, free tags and description, all written by whoever watched the video. It is the base, and it is deliberately human: we discarded measuring the files automatically, because whoever watched the film knows more than a colour histogram.

2. 02
   The published evidence, with its grade in view

   Each area has a rule written from the literature, independent studies and RVer’s own, and every suggestion shows its evidence statement with its grade: meta-analysis, RCT, review, pre-post or single arm. Where there is no decent literature, the suggestion says so rather than staying quiet.

3. 03
   What that department has learnt

   Without asking anyone anything: what the team picks, what it lets run to the end, what it cuts at two minutes, what it repeats for the same person. It moves the ranking by at most ±15%, only after five events, decaying to half in 120 days. A line that rose through use says so, so habit is never mistaken for evidence.

03Every module is asked along its own axis

The question changes with the module, because the clinical work changes too.

RVerby outcome

Anxiety and stress · pain · palliative care · sleep · memory · cognitive stimulation · apathy and motivation · orientation. Returns scenarios from the library.

RVer Motionby region and function

Shoulder and arm · hand and wrist · hip and leg · foot and ankle · balance and gait · strength and endurance · spine and trunk · neck · breathing · dual task. Returns exercises and activities, with a fit bar: how targeted that item is to what was asked.

RVer Neuroby cognitive domain

Attention · memory · visuospatial · executive function · speed. Returns tests.

RVer Exposurehas no assistant

And that is a decision, not a gap: the phobia is in the content’s name, the clinician picks by it, and an assistant would add nothing.

RVer Motion, RVer Neuro and RVer Exposure are modules in development, not covered by the Class I medical device registration (Infarmed), which covers the base product.

04What RVer AI does not do

• It does not diagnose, prescribe or assess

  It ranks content against a goal the clinician chose. No number it shows is a cut-off or a norm.

• It does not replace clinical judgement

  It assists. The choice is always the team’s, and the next level only runs if someone picks it.

• It does not leave the device

  What one department learns stays in that department. Nothing is pooled across clients, nothing trains anything central, there is no cloud and no account (GDPR).

• It does not promise where evidence is weak

  Where the literature does not support it, the suggestion shows the low grade instead of hiding it. In areas without decent evidence, it says so in full.

05Where it stands today