Free Humanoid Corpus

A structured prior art commons covering robots, humanoids, and humanoid-adjacent entities from private companies, open source repositories, science fiction (film, TV, comics, games, literature), and academia.

The corpus IS the prior art commons. Every entry, at the moment of timestamped commit and quarterly release, is a defensive publication. There is no separate commons layer above this catalog — the structured data here, combined with the cryptographic timestamping in releases/, is itself the invalidity-contention material.

License

CC0-1.0. Public domain dedication. See LICENSE.

What's in here

SCHEMA.md           — schema spec (currently v0.2)
PLAN.md             — multi-phase build plan
TIMESTAMPING.md     — release ceremony for cryptographic timestamping
MIGRATION_REPORT.md — v1 to v2 migration notes
README.md           — this file

corpus.jsonl        — master corpus (one JSON object per line)
private.jsonl       — generated mirror, entries with corpus="private"
open.jsonl          — generated mirror
fictional.jsonl     — generated mirror
academic.jsonl      — generated mirror

CORPUS_INDEX.md            — generated alphabetical index
lineage.json        — generated ancestor/descendant DAG

cross_cuts/         — generated per-subsystem prior art views
  CORPUS_INDEX.md          — table of all cross-cuts with counts
  <tag>.md          — chronological view of all entries disclosing <tag>

tools/
  validate.py       — schema and quality-bar enforcement
  index.py          — generates CORPUS_INDEX.md, lineage.json, per-corpus mirrors
  cross_cuts.py     — generates cross_cuts/

release.sh          — quarterly release ceremony with timestamping

Quick start

# Validate the corpus passes structural and quality-bar checks
python3 tools/validate.py corpus.jsonl --strict

# Regenerate all derived artifacts after editing corpus.jsonl
python3 tools/index.py .
python3 tools/cross_cuts.py .

# Run a quarterly release with cryptographic timestamping
./release.sh 2026.Q2

Current state

As of latest commit:

• 56 entries, 41 commons-grade and 15 draft
• 48 subsystem cross-cuts generated automatically
• Lineage DAG with 0 cycles, 18 to-do items (referenced-but-missing entries)
• Date range: 1920 (R.U.R.) to 2024 (multiple)

Quality bar

An entry is commons-grade (not draft: true) when:

1. disclosure_citation resolves to a primary source verifiable by a third party.
2. first_disclosure_date is the earliest verifiable public disclosure, defensible against challenge.
3. prior_art_notes reads as a 102/103 anticipation analysis someone unrelated to us could cite without rewriting.
4. sources cite primary references.
5. For patented entries, ip_citations lists actual patent numbers.

Entries that don't yet clear this bar are merged with draft: true and have explicit work items in their notes field.

Cross-cuts as the working tool

The cross-cut files (cross_cuts/<tag>.md) are the operational form of the commons. Each file lists every entry disclosing a given subsystem, in chronological order, with their prior_art_notes and disclosure_citation inline.

Use these when assessing patent claims:

• A claim about anthropomorphic robotic hands? Open mechanism-anthropomorphic-hand.md and read 20 disclosures starting at WABOT-1 (1973).
• A claim about quasi-direct-drive actuators? Open actuator-electric-quasi-direct-drive.md and read the chain anchored at MIT Mini Cheetah (2019).
• A claim about safety supervisors? Open safety-hard-constraint.md and read 10 disclosures starting at Asimov (1940), through Sha's Simplex (1995), HJ reachability (2005), CBFs (2007), ISO 10218 (2006), shielded RL (2018).

Contributing

Contributions are welcome under CC0-1.0. By submitting an entry you dedicate the contribution to the public domain.

A good entry is one that holds up as defensive publication. The contribution guide (see PLAN.md Phase 2) covers what that means in practice. Briefly:

• Cite primary sources, not Wikipedia.
• Identify specific subsystems disclosed using the taxonomy in SCHEMA.md.
• Write prior_art_notes as element-by-element anticipation analysis.
• For patented entries, point back at academic or fictional prior art that anticipates the claims.

If unsure, mark the entry draft: true and document the work needed to clear the quality bar.

Provenance

Originated within Open Interface Engineering (OpenIE) as part of the Free Humanoid project. The corpus is structurally separate from Free Humanoid: Free Humanoid uses the corpus, but the corpus serves the entire embodied robotics field, not any specific platform.