01
Now
Phase One software evidence
Compatibility, scorer and fusion implementation, measurement system, matched baselines.
ACTIVE / DESIGNED
Next unlockA reproducible result that justifies the next experiment.
Roadmap
A staged path, with gates before scale.
The roadmap makes dependencies visible. There are no percentage-complete bars because progress is defined by evidence, not appearance.
LAST UPDATED · AUGUST 23, 202602
Research extension layer
Formalize the mechanism set
Layered evaluation, hierarchy, packets, few-attempt comparison, two-direction learning, agent search.
UNDER FORMALIZATION
Next unlockBounded experiments with strong conventional baselines.
03
Next
Selective sensing
Wide-range force sensing and Stage-1 glove candidates only after software evidence identifies missing information.
POST-PHASE-ONE
Next unlockEach sensor earns retention through incremental value.
04
Then
Human physical data
Staged sensorized glove with synchronized contact, force, pose, motion, and outcome records.
POST-PHASE-ONE
Next unlockSafe, reliable, time-aligned collection.
05
Then
Robot-native data instrument
Execution and scientific data generation; investigate an active or shape-changing palm.
FUTURE ROADMAP
Next unlockA measured data advantage, not a uniqueness claim.
06
Then
Correction loop
Human correction, retargeting, safety, labeled corrective examples, and outcome tracking.
FUTURE / CONDITIONAL
Next unlockFeasibility and learning value must be validated.
07
Long term
Compounding training layer
Software evidence → sensing → proprietary data → models → robot data → correction → better models.
LONG-TERM
Next unlockEvery loop remains governed by evidence.
How the stages compound
Each stage inherits the evidence of the one before it.
Nothing on this roadmap starts from zero. Every stage carries forward the measurement rules, representations, and validated structure earned by the stage below it, and only widens the system when the evidence holds.