zenocognition
A sample day - the same surface that runs live on your own machine.
snapshot● sample · democlaude-opus-4-72026-06-16 20:45 UTC
Sample data. A grounded simulation of a representative day, shown so every panel is populated. The live dashboard runs on your own machine and reads only your local capture.Take the guided tour →
Attention lowHUD says “ride it out”score 25 is below the red threshold of 55 - step in or take a break.
01:30Same signal as the HUD
Cognition now
Tokens today
in / out / cache1.0M
input
43k
output
35k
cache rd
862k
cache wr
85k
Context window now
near the ceiling - compaction soon
Attention now
● lowsame signal as the HUD bar01:30
Focus over today
Attention timeline
48 samples today32 below threshold
The correlation
Token load vs attention, over today
tokens (left) attention (right) red threshold 55
Does load erode focus?
Tokens vs attention
Pearson r
-0.40
moderate inverse
Heavy token use rides with declining attention - the load-erodes-focus signal.
n = 48 paired samples
Window fill
Context usage today
Drill-down
Recent sessions
SessionRunsInterv.TokensAttn
Totals
Capture at a glance
Sessions
0
0 open
Agent runs
0
Interventions
0
Peak agents
0×
Accept rate
0%
Surveys
0
SCED 37/90
Cadence
Sessions captured per day
Oversight
Interventions by type
intervene855114ms
approve645097ms
redirect254985ms
rollback144631ms
Fleet
Agent runs by model
claude-opus-4-7116
claude-sonnet-4-689
claude-haiku-4-541
gpt-5-codex14
Concurrency
Peak agents per session
1× agents16
2× agents12
3× agents5
4× agents5
5× agents2
6× agents1
37/90
SCED
Local · this machine
Your idiographic corpus
Each zeno survey --autonomy adds a pre-registered single-case occasion toward N=90. Within-person, randomized, frozen schedule - the proof you live in.
Public · zeno.center
The RTLX-S research
The supervision-load instrument, discriminant-validity threat model, and the between-person whitepaper. Your local findings become the lived evidence behind the published claim.
zeno.center →