Eye strain signals: blink rate, eyes-closed, microsleeps
Blink rate alone is noisy. Eyes-closed seconds and microsleep events are stronger signals, especially when read together against your own baseline.
- Blink rate by itself is too noisy to act on; pair it with eyes-closed seconds.
- Microsleep events are rare but explain otherwise unexplained bad blocks.
- Visual strain signals get useful only when compared to your own same-hour baseline.
Eyes leak the signal before you feel it
Most fatigue tracking relies on what you tell it. Self-reports. Energy sliders. *How tired do you feel?* Useful, except that by the time you can answer honestly, the strain is usually already an hour deep.
Sarenica leans on the camera instead because the eyes drift first. Three signals do most of the work in this layer: blink rate, eyes-closed seconds per minute, and microsleep events. Each is noisy alone. Read together against your own baseline, they get useful fast.
Blink rate is the easiest signal to misread
A normal screen-facing blink rate sits around 12 to 18 per minute. Drop into deep focus and it can fall to single digits. That is not necessarily strain — it can be concentration.
Which is exactly why blink rate alone is a weak fatigue signal. It moves with focus, with task type, with light, even with how you sit. Sarenica records it because it is one input into a larger read, not because it is conclusive on its own.
- Blink rate alone → context layer, not a verdict.
- Low blink rate + high concentration → good focus.
- Low blink rate + rising eyes-closed seconds → the start of strain.
Eyes-closed seconds is the more honest signal
Eyes-closed seconds counts how many seconds in a minute your eyes were closed beyond a normal blink. Heavy blinks, long blinks, the small unconscious pauses that creep in when you are pushing through.
It is more honest than blink rate because it does not move much with normal focus. The long-blink pattern is a quieter and more reliable strain tell.
In your own data the number usually stays near zero through morning blocks, climbs in the afternoon, climbs again in late evening. The shape matters more than the absolute number, every time.
Microsleeps are rare. That is the point.
A microsleep event is a brief involuntary closure long enough that the system flags it as more than a blink. They are rare in normal work — you might see zero for days at a time.
When they show up, they almost always cluster around a specific hour or after a specific session length. That is the whole reason the signal exists. Not to count strain in real time, but to *retroactively explain a block that looked off* and you could not say why.
If a weekly report says your worst block was Wednesday at 14:00 and the microsleep events line shows a small spike there, you have a far stronger explanation than guessing.
How Sarenica combines them
No single eye signal triggers a fatigue call. Sarenica reads them as a small panel: blink rate as context, eyes-closed seconds as the burden line, microsleep events as the rare exception evidence.
They feed into a per-minute eye-strain score, which then feeds the larger fatigue and focus reads. That is why your weekly report can claim "30-45 minute sessions show 38% higher fatigue" without ever mentioning your eyes — the eye signal is folded in upstream.
What to do once a pattern repeats
Single days do not matter much. A repeating pattern across normal workdays does. If eyes-closed seconds reliably climbs after 11:00, the question is no longer "am I straining" — it is *"what changes if I cap demanding blocks at 30 minutes before noon."*
That is the kind of question your weekly report is built to answer. The eye signals are the why; the report is the so-what.
- 1Watch your weekly report for two consecutive weeks flagging the same pattern.
- 2Pick one block to cap at 30 minutes, in the hour where eyes-closed seconds climbs.
- 3Compare the next weekly report against the previous two — that is where the change-attribution layer earns its keep.
