Start with evidence
We separate documented prototype evidence, current engineering work, and longer-term research questions.
Research that keeps the evidence in view.
FSquare R&D explores persistent, evidence-grounded AI and specialised business models. We publish the direction clearly while keeping unverified claims and private implementation details out of the public record.
A public research view
01Observe the environment
02Keep source-linked memory
03Investigate uncertainty
04Verify before action
05Adapt with limits
Ambition without a black box.
Keep people in control
Permissions, evaluation, recovery, and human judgement remain part of the design problem.
Make it useful in context
The goal is not a clever prompt. It is capability that can remain useful across real documents, tools, projects, and change.
Aegis Humaniform.
A public-safe research program for AI that can remember, investigate, verify, act, and adapt across long-running work—not just respond to the latest prompt.
The research page separates observed evidence from work that still needs validation, so collaborators can understand both the direction and the limits.
Read the research recordA research question is a better start than a sales promise.
If you are exploring a serious AI problem, a specialised model, or a governed agent system, tell us what needs to be understood, tested, or made more reliable.
Start a research conversation