Open data and models
SpaceAI
The Australian corpus, the sp32k tokenizer, SpaceAI-0.1B and the Swag training system: what each is, what is public, and how it is licensed.
See SpaceAIResearch / evidence
The claims should be inspectable.
This is the home for experiments, implementation evidence and the limits of each conclusion. Published notes are exploratory. A dated provenance timeline and a reproducible evidence register are being assembled; they are not presented as complete.
Published material
Read the work that is available now.
Method and results, exploratory
Five Pillars of AI Safety
A small, self-run set of model-behaviour probes. Read its method and limitations before drawing broader conclusions. It is not an independent audit or certification.
Read the whitepaperField notes
Writing
Notes on model behaviour, failure modes and the decisions behind the build. These are working observations, not claims about every model or deployment.
Browse writingCurrent implementation
PulSen Craft
The product page describes PulSen Craft Desktop and PulSen Craft Fleet in the current private beta. Public technical demonstrations and a claim-by-claim evidence pack are still being prepared.
Explore current capabilityEvidence work in progress
What will make the story easier to verify.
01Founder provenanceDated primary artifacts before a public chronology is asserted.
02Capability registerClaim, implementation, evidence and demonstration method for each major feature.
03Reproducible experimentsMethods, data, results, limitations and safe reproduction notes.
Until those artifacts are ready, the site keeps the technical lineage concise and avoids claims of influence on other AI companies.
Research enquiry →