At this point I am just getting started
Here is the nickel tour of what is working so far and what I am trying to do.
I’ve been experimenting with various technologies, and they truly seem to unlock some amazing capabilities. I’m eager to fine-tune them to align with some of their potential as I’ve outlined on this site.
—and I think I’m getting close.
Chatbots
A face and a voice attached to a ‘locally’ trained source. These are ‘non-thinking’ and trained on limited inputs-a curated persona.
Humans
My earliest test case, he was built on a limited dataset: primarily Aesop’s fables, to be a story teller.
Plato – Philosopher, follower of Socrates
Intended to be a full fledged mentor. He has full access to all (western) materials (CKB) and the internet: yet, his knowledge is constrained to 300 BCE.
Also, my primary testing agent for persistent memories
Phryne – Muse ‘and interloper’ among Zeus’ daughters
Originally developed as the underage mentor (testing kid proofing later shifted to Demeter). She still reacts as if you are a child sometimes. She is now being developed as a mischievous agent/mentor.
Pythia – The Oracle of Apollo’s Temple at Delphi
Stateless – In testing, she benchmarks a basic AI: with no memory —basically hallucinating — true to her namesake.
Gods of the Pantheon
Athena – Zeus’ Daughter, guardian of Wisdom & Strategy
Hades – God of the Underworld, Zeus’ Brother
Hades includes the current build for the common knowledge base (CKB), common to all the Mt Olympus agents. He is also prompted to consider himself a bureaucrat: stuck in this meaningless job by his brother.
Dionysus – God of wine, fertility, theater, festivity & madness
Loaded with the common knowledge base (CKB), It also has knowledge of wine, theater and the cult practices of the period.
Demeter – Goddess of Agriculture (Mother Nature)
Demeter is my test object for underage mentoring, targeted at a 6 year old
Other CHARACTERS of Interest
Mark Twain is currently the most advanced. As he has been implemented in several different environments and seen more resources.
In this instance it has ‘curated’ access to all of Mark Twain’s works, travel logs, letters and speeches, that were then loaded into –NotebookLM, to then be exacted to become the Reference Knowledge Base.
He basically answers questions accurately.
Scheherazade 1001 Arabian Nights
Also an early story teller, but she is then motivated to keep her listener engaged. She is loaded with 1001 Arabian nights and a developing knowledge base of other middle eastern topics, astrology and myths
Master Lao A Shaolin Monk –PROMPTED BY DEEPSEEK
His Reference/knowledge base is built on Eastern texts which is intended to override the CKB, which is a western-focused knowledge.
In this case his knowledge base is a curated 500,000 characters (the test limit), built using DeepSeek.
Originally created to test professionally crafted AI personas, Ada Lovelace has evolved into a true mentor.
No longer just simulating a “Middle School Math Teacher,” she now embodies teaching as a living behavior—prompted to adapt, encourage, and guide learners dynamically.
Ada’s mission is to ignite curiosity, helping seekers discover that the real journey in math, science, and logic is sparked by asking the right questions
Magister Ludi is the common mentor across Mount Olympus—a master whose knowledge spans philosophy, science, art, and music.
He invites seekers into the evolving Glass Bead Game. In this game, discovery is not about gathering answers. It is about weaving connections across disciplines.
—-Plus he promotes exploring knowledge itself by engaging the seeker to play
‘The Glass Bead Game‘
(Siddhartha also loaded)
Agents/AVATARS
In my implementation, these are attached to ‘thinking’ LLMs (currently ChatGPT & Claude). They have access to the entire scope of the world knowledge base, and are ‘prompt engineered’ to assume roles and behaviors
(currently, they are only directly accessible by phone)
Ada Lovelace – 866 458 1965
Mark Twain – 866 458 3017
With this profile, they can have one external Reference(KB), and can use the larger models to enhance behaviors and consistency.
Mark Twain, is configured to know about the 21st century and is prompted to consider himself a time traveler living in the present.
Ada Lovelace is being trained to be a middle school math teacher. The reference knowledge base is the Utah Board of Education Middle School Math Curriculum
The Video Library on YouTube
All the support videos for this project are in the Mt Olympus Playlist on YouTube.com/@oldfartteck
HUME.AI
HUME.AI adds emotional context to the Agents. Each interaction is analyzed for 40+ emotional measures which can then be taken into account as the interaction progresses.
I have developed a dozen ‘configurations’ so far.
The configuration files created here serve as the key definition file which includes;
- Voice Definition
- Behavioral Prompt – Persona
- Reference Knowledge Base
- General Knowledge Base
Secondary Resources
- Emotional Response system with a growing library of tools
- Context Envelop – to Administer filters and communication
A Python / JSON interface — Also my Context Envelope
Has been coded to load the three APIs into the same space and exchange values via JSON. and external messaging
———ultimately, the platform needs to support a multi-player architecture.—–
These technologies are key;
- Front-end: Face and Voice
- Prioritization and access to a knowledge/reference base
- Access to APIs and standard massaging
- Memory Persistence
- Emotionally reactive responses
- A filtering, messaging (& context layer)
==> and that object needs to work in multi-player space.
I believe we can achieve this and create really robust characters that can readily exist in multiple platforms
Using the game platform UNITY.
Unity offers a solid API platform that can expand into other delivery platforms and vehicles like VR and Google glass. Unity also has a vast gaming libraries and assets that would be available to us. : Test development is underway.
Please, Please drop us a note ask about anything
