Introducing NeoThymos

Welcome to NeoThymos!

We are glad you took the time to visit and check us out.

This post is pre-release, pre-Beta in fact, so the design of NeoThymos is still somewhat fluid. The system and this post may yet change. Bear with us as we finalize preparations for release.

First - hi, I’m Paul Pakozdi the principal developer of NeoThymos. This first post will be a little bit about a lot of things but will hopefully include enough about some things to satisfy your curiosity and encourage you to give NeoThymos a try. Future posts should be a bit more focused but today we are going to wander a bit. Some historical context, a problem in agent development that needs attention, a bit of psychology, a brief introduction into how NeoThymos can help, and a smidgen of how we validate the fidelity of NT Personalities. So let’s begin.

It’s the summer of 2026. Modern AI is out of the box in a big way. Having followed AI for much of my life and having worked with GOFAI inside of a large computer company before the turn of the century, I am amazed at where we find ourselves today. AI today is what we would have called “Connectionist AI” back in the 1990s. In the time since, the focus has shifted from backward and forward chaining rule based systems to neural networks made up of tens of trillions of connected weights.

I’ll refrain from reminiscing further today and simply say that it is incredibly humbling to look back at the tremendous progress made by researchers in the field of AI since the 1950s but especially the progress of modern AI this past decade. In a very short time this research has changed the world forever even if most of the world does not yet grasp the magnitude of that change.

Today the software industry is frantically focused on agentic systems; deterministic harnesses using stochastic systems (LLMs), both simple and complex. Such neuro-symbolic compound systems are designed to get things done and in a myriad of configurations; individually, in groups, in swarms. Some of these agentic systems are accessed through control surfaces designed for scientists and engineers. But many require a surface suitable for non-technical human interaction. Even groups of agents organized hierarchically often have an agent whose responsibility is to communicate with a human owner so that it can orchestrate other agents to get things done.

If you have worked for any time with LLMs from multiple vendors or even with models of the same lineage you know they all tend to “feel” different in conversation. (Who can forget the OpenAI ChatGPT v4.0 controversy).

If you are developing AI agents you know that every AI agent ships with capabilities. Tools, memory, skills, guardrails, the checklist grows every month. But ask what an agent is like, how it meets a stranger, what it finds funny, where its patience runs out, what it refuses to become, the answer is usually a paragraph of adjectives somebody typed into a system prompt on a Tuesday night. It is a thoughtful ad hoc construction at best, or often as not, a confused and conflicting word salad.

We think that’s backwards. Personality isn’t a coat of paint you brush over a capable system. It’s structure — and it deserves real engineering.

Grounded

At its very heart, NeoThymos uses a personality model rooted in the psychometric tradition, research psychology rather than vibes. We have chosen the the HEXACO model to represent an agents ‘core personality’. This model goes beyond the OCEAN and Big-5 models of personality, refactoring the old domains to add the new domain of Honesty-Humility (what we refer to as Integrity in NeoThymos). We use other psychological models deep inside of the product along with rules derived from modern psychological and psychometric research to project this core personality onto other psychological vector spaces that are useful in defining a full personality profile.

No magic, just a firm grounding in the collected research of the human psychology of personality, emotion and moral reasoning.

Synthetic

NeoThymos is an engineered approach to building personalities outside of human minds. It builds personalities from the ground up. The HEXACO model provides the surface that developers work with to configure a core personality. The system then deterministically projects the core personality on to other psychological models using rules from established psychological research to create a full Personality Profile. It starts with a with a 25-dimension core personality

From that core, our projection engine derives a complete psychological profile: baseline emotional disposition, sensitivity, moral center, temperament and communication style boundaries. Change one facet — make her a little bolder, make him slower to forgive — and everything downstream shifts coherently, the way it does in a person.

No adjective soup. No “be helpful and slightly sarcastic.” A personality with internal mechanics — one that holds together under pressure, because it was derived, not described.

Portable

Here’s the part we care about most: the personality is yours, and it travels. When you finish defining a Persona in NeoThymos, you don’t get a vendor lock — you get a NT Persona Kit: a self-contained package your chosen agent framework
reads to become that character. Design once; deploy to Hermes, Open Claw, NanoClaw or a hand crafted agent harness. Then take it with you when you move or begin testing other frameworks.

---

We call the product category Synthetic Personalities — personalities grounded in real world psychological models, synthesized from engineered structure and portable across agentic frameworks.

As far as we know, we’re the first to make this particular kind of tool generally available to the development community.

What’s live today

The NeoThymos studio has four rooms.

• The Gallery is the main workbench: a place where personas that are under active development reside. Its a shelf of characters you’re building or have adopted from our archetype catalog, a catalog of pre-defined personality types.

• The Library is where you store them: an archive for personas. Each month you can move a persona from each slot in your Gallery to the Library, freeing up that slot for the creation of a new persona. Personas in the Library can be redeployed at any time or be moved

• Staging is where you prepare your persona to go out into the world. This is where you manage a persona’s lifecycle through editing, synthesizing, locking and deployment of a persona to an AI agent runtime environment.

When a persona is ready you lock it — a commitment moment where its portrait and character description are synthesized and its identity freezes — and from there you deploy Kits to your target framework(s).

We’re currently in an invite-only preview. [How to request access / waitlist link.]

Where this goes from here

New Domains Agentic computing is the first domain, not the last. The same grounded core that drives an agent can drive a novel’s character bible, a TTRPG party’s NPCs, a game studio’s cast. Those workshops are on our roadmap, and the personalities you build today will carry into them — that’s what portable means.

Together We Create Bespoke Personalities. We’d love to show you. Take one of our pre-defined Archetype Persona for a free spin or sign up

[Signature / contact / follow links.]