JAKASII research

Research in progress.

Our research begins with questions rather than conclusions. We are studying A·we, AI safety, cybersecurity, agents, recursive self-improvement, and reinforcement learning as areas of ongoing work.

A programme of questions.

JAKASII is developing a research programme around language, reliable systems, and learning. The work is organised around questions that can be examined carefully rather than claims made before the evidence is ready.

A·we is the first language focus. We are studying how a language programme can respect variation, preserve meaning, and remain answerable to the people and places that give the language its life.

AI safety and cybersecurity shape the technical side of this work. We are examining how systems can state their limits clearly, protect important boundaries, and leave decisions open to review.

We are also researching agents: systems that can support practical work through defined tasks, clear constraints, and accountable outcomes. The question is not simply what an agent can do, but how its work can remain understandable and controllable.

Reinforcement learning (RL) is part of that inquiry. We are studying how feedback can guide learning while goals, safeguards, and evaluation remain explicit.

Recursive self-improvement (RSI) is a longer-term research direction. We are interested in how systems may improve their own processes through deliberate, bounded changes that can be checked before they are relied upon. Together, these areas inform the development of JAKASII’s own base model.