Deployment signal
A named program, agreement, test or benchmark can enter the directory when a dated original source confirms that it exists and shows its current stage. Planned work remains clearly labelled as planned.
Editorial policy / Version 1 / Updated 2026-09-04
How information
gets onto the site.
This policy explains how HumanoidUptime finds, selects, checks and corrects information. The methodology separately explains the reliability fields and source levels.
01 / Scope and discovery
What we look for—and where we stop
- What we cover
- Publicly reported humanoid deployments, pilots, operator programs and physical tests that reveal something useful about real operation. A product announcement may appear as a signal, but it is not enough for a full Reliability Record.
- Where we search
- We start with publications from operators, manufacturers, regulators, investors, research institutions and test bodies. News coverage can point us in the right direction, but the original source remains the basis for a record whenever it is available.
- Languages
- English and German are currently covered most consistently. Sources in other languages are included when their wording and origin can be checked. The site does not yet claim complete multilingual coverage.
- What “not disclosed” means
- Every record shows when it was last checked. “Not disclosed” means we could not find a usable figure in the sources listed on the page by that date. It does not mean that no private data exists.
02 / What gets published
Two ways information appears
Full Reliability Record
A full record must clearly identify the robot or configuration, operator or test body, task, place and phase. All nine reliability questions are then answered or marked as unknown, with a source attached to every important claim.
03 / Review workflow
Seven checks before publication
- 01
Discover
Find a candidate and identify the closest available primary sources.
- 02
Set the boundaries
Keep model generations, sites, tasks, time periods, fleet claims and future plans separate.
- 03
Separate the claims
Check each important statement on its own instead of accepting an entire release as one fact.
- 04
Label the source
Show whether a claim was independently measured, confirmed by an operator, reported by a manufacturer or presented only as a capability.
- 05
Show what is missing
Answer all nine reliability questions without turning an unknown value into zero, failure or an estimate.
- 06
Review
Before a record is published, a person checks it against the sources linked on the page. Not every record has a second or external reviewer.
- 07
Version
Publish the review date and preserve material changes, corrections and evidence transitions instead of silently replacing history.
Automation and AI
Assistance is not evidence.
Automation and AI may help find, organize, compare and draft material. They are never treated as evidence and cannot make a source more reliable than it is. Automated discovery does not publish records, invent missing fields or decide that two measurements are comparable.
Exclusions
A weak signal stays weak.
Rumours, unattributed social posts, promotional video alone, generic specifications and future agreements without a bounded result cannot establish field reliability. They may only guide further research or remain a clearly labelled signal.
Corrections and disagreement
Better evidence wins.
Anyone may submit a correction. The submission is tested against the same source and context rules. A covered organization may correct facts but receives no veto over evidence classification. Material corrections create a visible change entry and, where applicable, a new record version.
Submit a correction →Coverage
The site does not cover everything—yet.
The directory is growing in a deliberate order. A company appears because its deployment is relevant and has sources worth checking—not because HumanoidUptime endorses it or considers it a market leader. The numbers on the site describe the material reviewed here, not the entire humanoid market.
How the numbers are judged
Selection is only
the first step.
The methodology explains what each reliability field means, how sources are labelled and when two numbers can—and cannot—be compared.
Read the methodology