Methodology

How we score a humanoid

Eight capability axes, broken into 40 sub-axes, each anchored 0–5 and tied to evidence. The Cyborgs Octant — explained in full.

The pipeline · 5 steps
Step 1
Capture
Crawl vendor + analyst sources daily.
Step 2
Research
Analyst deep research per offering, human-reviewed.
Step 3
Score
Each sub-axis anchored 0–5 against the rubric.
Step 4
Calibrate
Cross-offering panel reviews every axis together.
Step 5
Publish
Scores flow to the Index, Compare and offering reports.

Live scores today are produced by our v2 pilot-gated engine — the raw 8-axis read, capped by commercial-scale evidence. A v3.1 proof score engine is in pilot; we will publish its first calibrated scores and full mechanics together, not separately.

The Cyborgs Octant

8 axes · 40 sub-axes

Specs are not capabilities. Demos are not deployments. Every sub-axis below has a 0–5 anchor descriptor — the exact words our analysts must see in evidence before we award the score.

Flat-ground walkingFlat-ground walkingUneven terrain & slopesUneven terrain & slopesStairsStairsDisturbance recoveryDisturbance recoverySpeed & enduranceSpeed & enduranceGrasp varietyBimanual coordinationForce controlIn-hand reorientationTool useSensor stackScene understandingSpatial memoryObject permanenceSemantic world modelAutonomy levelTask horizonGeneralizationInstruction followingRecovery from failureImitation learningSim-to-realVLA / foundation policyData flywheelTransfer & fine-tuneSpeech I/ONatural-language instructionIntent recognitionSafe proximity behaviorSocial presentationNamed pilotsField hoursCustomer diversitySupportabilityCommercial postureFunctional safetyCertificationsE-stop & failsafeIncident transparencyForce / impact limitsLocomotionManipulationPerceptionAutonomyLearningHuman–RobotDeploymentSafetyCyborgs OctantLocomotion5 sub-axes
Active axisSub-axis← → cycle · hover to select
Locomotion
Flat-ground walking

Steady gait on level surfaces with payload.

weight 1.00·key flat_ground_walking
Scoring detail

Anchor descriptors for this sub-axis are published with the first Cyborgs Index release on 31 July 2026.

Evidence labels

Every datapoint carries a label

The UI never visually equates a vendor claim with an independently validated observation. Eight labels, applied at the row level.

unknown
We have no evidence on this question yet.
vendor claimed
The vendor says so. We have not seen it.
reported
Reported by trade press rewriting the vendor.
demo shown
Shown in a vendor or staged demo.
provisional
Independent but partial / single observation.
validated
Independently verifiable, replayable, or measured.
disputed
Sources disagree; we hold the score down.
analyst derived
Our analysts inferred this from other evidence.
Commercial Proof Map

Six stages from claim to scale

Readiness lives on a runway, not a colour. Position carries the meaning. An axis cannot exceed its commercial-proof ceiling — that's what pilot-gated means.

  1. Stage 1
    Claim
  2. Stage 2
    Demo
  3. Stage 3
    Pilot
  4. Stage 4
    Named deployment
  5. Stage 5
    Measured operation
  6. Stage 6
    Scale
Editorial independence

Sponsors and partners never influence rankings, evidence labels, deployment validation, or the methodology. We will refuse a sponsor before we will bend the rubric.

Change log

Methodology revisions are appended with each monthly Index release. We do not silently rewrite history — if we change a score or revise a deployment, the change is logged.