HumanoidUptime
Last evidence review · 04 Sep 2026

Public change history

The record should show
how it changed.

Material additions, corrections and methodology changes are logged here.

v0.9.0

Digit factory-use evidence

  • Published the Digit × Schaeffler Cheraw evidence record from operator, manufacturer and independent-reporting source trails.
  • Separated a more-than-one-year calendar evidence span and reported daily work pattern from cumulative runtime and measured availability.
  • Added the named Cheraw workflow to the deployment directory without inferring a delivered global fleet from the purchase agreement.
  • Expanded the corpus to eight records, 21 deployment signals and 72 reviewed reliability fields.
v0.8.1

Evidence language refinement

  • Introduced a consistent five-mark visual language for independent, operator, manufacturer, capability and undisclosed evidence states.
  • Made undisclosed fields deliberate findings through explicit labels and a restrained missing-data pattern instead of leaving ambiguous blank states.
  • Added accessible evidence-level text for disclosure-matrix and record-detail cells.
  • Standardized public record labels on the durable HU-R identifier format.
v0.8.0

Versioned editorial data foundation

  • Introduced PostgreSQL as the canonical editorial store while retaining deterministic JSON snapshots for the static build.
  • Added immutable record and dataset versions, field-level change events and source-observation snapshots.
  • Assigned durable record and claim identifiers without changing the published evidence conclusions.
  • Kept the public website statically served and independent from database availability.
v0.7.1

Legacy record normalization

  • Normalized records 001–005 to the provenance and denominator standard used by the newest audits.
  • Exposed the unresolved ten-versus-eleven-month scope difference in the Figure 02 × BMW public record.
  • Separated GXO operator evidence and the joint commercial announcement from Agility-only output, duration and autonomy claims.
  • Distinguished Walker S2 battery-swap autonomy from task autonomy and Figure 03 project language from measured field operation.
  • Reclassified the Unitree G1 1:49-hour result as scenario-derived endurance and added Fraunhofer's collision-test repeat context.
v0.7.0

Historical retail pilot

  • Published the Sanctuary AI × Canadian Tire/Mark's retail-pilot evidence record.
  • Preserved 110 correctly completed tasks as manufacturer-reported output without inventing a success rate.
  • Kept the pilot's robot generation and autonomy mode unknown rather than retroactively assigning the Phoenix name.
  • Expanded the deployment directory to 20 signals and the disclosure matrix to 63 reviewed fields.
v0.6.0

Factory POC evidence

  • Published the HMND 01 Alpha × Siemens evidence record from three first-party source trails.
  • Separated the reported eight-plus uptime hours from availability, MTBF and continuous autonomous runtime.
  • Added operator-corroborated throughput and autonomous pick-and-place success with explicit denominator gaps.
  • Expanded the deployment directory to 19 signals and the disclosure matrix to 54 reviewed fields.
  • Established an internal, validated 30-record research pipeline.
v0.5.0

Evidence depth

  • Added an 18-signal deployment evidence directory with explicit field, pilot, planned and research stages.
  • Added Established, Claimed and Still unknown verdicts to every full evidence record.
  • Published a strict comparison gate with concrete examples of misleading cross-model comparisons.
  • Added canonical URLs, JSON-LD metadata and an RSS evidence feed.
  • Extended source validation to the deployment directory.
v0.4.1

Public crawler access

  • Removed the site-wide noindex directives and crawler block.
  • Allowed OAI-SearchBot, user-initiated ChatGPT access and ordinary search crawlers.
  • Kept GPTBot blocked as an explicit opt-out from possible model-training crawls.
  • Published a sitemap covering every public HTML page.
v0.4.0

Independent evidence

  • Added the Unitree G1 EDU-4 × Fraunhofer IPA independent benchmark record.
  • Published independently measured energy, endurance, movement, safety, thermal and security findings with test context.
  • Separated per-charge endurance from cumulative runtime as a ninth disclosure field.
  • Added a data-driven Latest Evidence log to the homepage.
  • Introduced an internal risk-based review schedule and monitoring workflow.
v0.3.0

Focused evidence navigation

  • Moved evidence records directly below the homepage summary.
  • Reduced record cards to scannable previews that lead into the full audit.
  • Prioritized Records, Compare, Methodology and Submit evidence in the main navigation.
  • Added an accessible mobile navigation and clearer matrix scroll guidance.
  • Condensed the homepage methodology section and moved the changelog to the footer.
v0.2.1

Internal data boundary

  • Removed the public raw-JSON output and all links to it.
  • Kept the structured record set as an internal editorial source.
  • Deferred machine access to a purpose-built, versioned API.
v0.2.0

Evidence Edition

  • Added structured records as the canonical data source.
  • Published four detailed case files with claims, timelines and source ledgers.
  • Added the Figure 03 × BMW deployment watch record.
  • Published the full evidence methodology and public change history.
  • Added source-link validation and an evidence-submission workflow.
v0.1.0

Initial public evidence board

  • Launched the landing page with three initial evidence records.
  • Introduced the eight-field disclosure matrix.
  • Established the rule: We don't score what nobody has measured.