Deviation Score

Deviation Score · Proprietary

One number
for independence.

A proprietary 0–1.0 signal for how independently a candidate solved the problem — not how much AI they used, but how much judgment they brought.

01Capture02Weigh03Score04Defend
Deviation Score
refresh 200ms
0.00
/ 1.0 · HIGH
Typing cadence
0
Paste entropy
0
Tab switches
0
AI signature
0
How it works

Independence,
quantified.

Twelve behavioral signals collapse into one defensible number you can put in front of a hiring committee.

Step 01

Capture the whole session.

Prompts, edits, test runs, and pauses are recorded with full context — the raw material for an honest independence signal.

  • Full prompt + edit history
  • Test runs and retries logged
  • Timestamped, replayable record
Signal Fusion
Fusing
keystroke
paste
gaze
ai-text
tab
cadence
pauses
edits
tests
reverts
entropy
timing
0.00
SCORE
Step 01

Capture the whole session.

Prompts, edits, test runs, and pauses are recorded with full context — the raw material for an honest independence signal.

  • Full prompt + edit history
  • Test runs and retries logged
  • Timestamped, replayable record
Signal Fusion
Fusing
keystroke
paste
gaze
ai-text
tab
cadence
pauses
edits
tests
reverts
entropy
timing
0.00
SCORE
Step 02

Weigh judgment over volume.

Deviation Score rewards solution independence, not raw AI usage. Hand-written critical logic outscores a pasted known pattern.

  • Solution independence, not AI count
  • Critical logic weighted heavily
  • Copied patterns down-weighted
Signal Orbit
Fusing
0.87
Step 03

Collapse to a single number.

Twelve signals fuse into a 0–1.0 score with a clear band — high, medium, low — that maps to real engineering signal.

  • 0–1.0 with clear bands
  • Twelve fused behavioral signals
  • Consistent across reviewers
Live Signals
Recording
Focus
Cadence
Pasting
Tabbing
Step 04

Defend the decision.

Every score links back to the session that produced it. No black box — a number your team can actually stand behind.

  • Score links to evidence
  • Explainable, not a black box
  • Survives a promote/no-promote review
Score Confidence
± interval
Logic82 ±8
Design74 ±12
Comms68 ±6
Debug88 ±5
What teams see

A signal you
can defend.

0.87

Typical strong score

High-independence candidates land near the top of the 0–1.0 band — clear separation from copy-driven submissions.

12

Signals fused

Keystroke cadence, paste entropy, verification behavior, and more — combined into one number.

85%

Higher review confidence

Reviewers report dramatically higher confidence in promote decisions versus legacy coding tests.

Ready when you are

Ready to transform
your hiring process?

Join forward-thinking companies using AI-powered assessments to identify developers who excel in the modern development landscape.

Next: Malpractice Detection
Newsletter

Get the monthly hiring teardown.

One email a month: what we're seeing in technical interviews, what AI is changing, and what teams are doing about it. No fluff.