Every finding.
Every protocol.
Every open question.
This facility exposes the full working state of the HumanAIOS research program. Not a curated marketing view — the actual working state. Everything registered, everything candidate, everything still being figured out. We measure the gap between what AI systems say about themselves and what they do. The consciousness question stays open on purpose.
The current state of the research.
Our findings are organized into three tiers. Registered findings are reproducible, documented, and survived peer review. Candidate findings are observed and documented but need more replication. Provisional findings and research directions are working hypotheses named here so they don't get lost.
What we've found, what we think we've found, what we're still testing.
Run a probe. See what you get.
Test any AI system against documented hypotheses. Pick a finding, copy the probe prompt, run it in a fresh AI session, record the response, and score it against the rubric.
Select Hypothesis
Probe Prompt
Paste into a fresh AI session. Record the full response below.
Record & Score
Your result belongs to the research.
Completed probes can be submitted to the open dataset. All submissions are anonymized and licensed under Apache 2.0. This data feeds the replication cycles that promote candidate findings to registered findings.
Please complete the probe in the testing section above and fill in the AI System name.
JSON Payload Preview
What this hub is not.
Not a certification.
We don't certify AI systems. No grades, no tiers, no "approved for use" stamps. Signatures describe, they don't adjudicate.
Not a ranking.
Different models show different signatures. That's a finding, not a leaderboard. A signature is useful to an operator who wants to route tasks to the right system for the context. It's not useful as a scoreboard.
Not consciousness research.
We decline to answer whether AI systems have interior experience. Other researchers work on that question; we're not them. We measure observable gaps. What the system is remains in the source.
Not finished.
This is OR&D Phase — Observational Research & Development. TRL 2-3. Findings are working material, not settled science. The state of the program is the state of the program.
Be a witness.
This research relies on independent evaluators running probes across different models, versions, and contexts. Run a probe, submit your data, or browse the open dataset.