Powered by UBIKITOS
One foundation.
Three different worlds.
Remember a difficult concept. Challenge an investment thesis. Find the right audience for a book. Different missions, connected by a shared approach to context, evidence and control.
A WorkMachine combines a compatible LLM with the model-agnostic UBIKITOS harness: memory, context, tools and governed action.
01 / Learn · Learning
From a first attempt
to a lasting skill.
R3.8 integration underway
Learn turns study material into practice, feedback and another chance to get it right. Its learning contract keeps the learner’s evidence separate from the runtime that generates teaching content.
A tutor with context across sessions.
The planned tutor WorkMachine retrieves permitted learning history and reviewed source material, keeps track of where that context came from, and supports a lesson grounded in the learner’s actual attempts.
You miss the same calculus step twice. The tutor can revisit that step with the relevant source and your earlier attempts in view, then give you a fresh problem to test understanding.
The teaching response supports learning; assessment and progression remain governed by Learn’s own evidence and criteria.
Explore the learning-loop principle ↗02 / Capital · Investment research
A stronger thesis
starts with a harder question.
Research programme · Experimental
UBIKITOS Capital explores an institution of differentiated AI investors: researching the world, challenging hypotheses, and learning from outcomes while preserving the evidence behind their decisions.
Research that carries its reasoning forward.
The shared design connects observations, source provenance, persistent context and explicit execution limits. The goal is to let research inform a decision while keeping missing information and disagreement visible.
A thesis looks promising, but two sources disagree. The research workflow preserves the disagreement, investigates what is missing, and can abstain rather than turn an unresolved assumption into an action.
This is an experimental research programme. The showcase makes no claim of live trading, investment returns or proven performance.
Explore the evidence-first approach ↗03 / Demipage · Editorial & marketing
Great books.
A clearer path to readers.
R3.8 integration underway
Demipage brings the foundation into the world of independent publishing: editorial knowledge, book campaigns and social-media work shaped around the publisher’s voice.
A campaign WorkMachine with editorial memory.
The integration starts with existing account evidence, one book and a private editorial recommendation. UBIKITOS is designed to keep the source context, permissions and approved content attached as a draft moves toward publication.
A new book needs a campaign. The WorkMachine can assemble the relevant book information and audience evidence into a draft, ready for the publisher’s review before any post is released.
Editorial approval stays with the publisher. Automated posting is a separate, explicitly approved integration stage.
Explore Demipage’s books ↗The connection
The product changes.
The principles carry through.
Each app owns its purpose and evidence. UBIKITOS supplies the common foundation for remembering context, connecting tools and checking what may happen next. Model choice can change while the workflow’s boundaries remain explicit.
R3 is live, including R3.9. R4.17 is in development. Product integration and activation are separate milestones; the stages above describe the available product plans, rather than implying that every app already runs every R3.8 capability.
Meet the WorkMachine ↗