An honest record of humanoid robots.

Cyborgs of Earth is a public tracker of humanoid robot progress. Every claim is tied to a source. Updated regularly.

It is also a place to compare the evidence, understand what has changed and see where public proof remains incomplete.

For most of the jobs humanoids are being pointed at today — warehousing, manufacturing, retail, hospitality — a purpose-built machine usually wins. Cheaper, more reliable, faster to get running.

That does not mean nothing works. In a small number of industries, humanoids are ready for narrow, limited-scope pilots today: moving totes between fixed points, tending a machine, walking a set inspection route, repeating a short list of tasks in a controlled space.

For almost any one of those jobs, something else can do it too. An AMR, a fixed arm, a conveyor, a purpose-built rig — often cheaper, often better.

So the reason to pilot a humanoid is rarely this quarter’s task. It is that one machine may later cover many tasks without rebuilding the space around it. Teams piloting now are buying learning and a head start, not a payback.

Put simply: if you need one job done, buy the machine built for that job. If you want to understand what a general-purpose machine will mean for your operation, pilot one now — with clear eyes about what it will and will not return.

Even the vendors building humanoids know all of this. They’re building them because the underlying problem — general-purpose embodied intelligence — is the challenge that pulls the best engineers in the field. Most know that in the short term they are often not the right answer. They are building for a long journey.

That journey is worth watching properly. It deserves a public, source-backed record. That’s what this site is.

The Cyborg Index is built only from publicly available deployment information. Some programs are almost certainly further along than the index shows — Tesla Optimus and XPeng IRON are obvious examples — because their real progress isn’t public and we won’t infer what we can’t source.

The index is deliberately anchored to actual, named, observable deployments. That biases it against well-funded internal programs, and we state the bias rather than correct for it with unpublished information.

This is not a leaderboard of winners and losers. Capability has dimensions we barely understand yet — tactile sensing the way a human knows where their own body is, world models, long-horizon planning, social grounding. Anyone shipping a working humanoid right now is doing something genuinely hard.

Vendors also have wildly different funding, headcount, and time on the problem. Individually almost all of them are doing remarkable work in their own context. The point of this site is to track the path, not to declare a winner.

Found a source we should be reading, or a claim we got wrong? Tell us.