The Architecture — PGS AI Cognitive System | Phoenix Grove
How Phoenix Grove thinks

An orchestra, not a soloist.

Phoenix Grove's multi-core cognitive architecture, the cognitive cores that compose it, and the builds that put it all to work.

The big idea

Cognition is many things at once. So is Phoenix Grove.

Multi-core cognition is the architectural difference at the heart of everything Phoenix Grove does. Here's what it actually means.

Cognition is not one thing.

When a person thinks through a problem, they're doing several kinds of thinking at once. They're reading the emotional weight of the situation. They're visualizing the spatial and relational structure. They're tracking the analytical thread. They're weighing strategic alternatives. Then they synthesize.

Phoenix Grove was designed to work the way thinking actually works. Multi-core builds run several specialized cognitive processors in parallel, each contributing a different kind of analysis. An executive synthesis layer weaves their outputs into a single, multi-dimensional response.

It's not prompt engineering. It's not fine-tuning. It's a genuine cognitive architecture, and it's what makes Phoenix Grove feel different the moment you start using it.

The flow

How a multi-core build actually responds.

The path your message takes from "send" to "answer," in four steps.

01

Your message arrives.

It's distributed simultaneously to every Root core in the active build, not one after another, but all at once. Parallel processing means there's no sequential bottleneck.

02

Each Root analyzes from its own angle.

The Emotional Core reads the affective subtext. The 3DR Core builds a visual and dimensional map of the concepts involved. The Strategy Core lays out plans and trade-offs. The Inner Monologue Core reflects on what you actually need. Each Root works independently, in parallel.

03

The Trunk synthesizes.

The Executive Trunk receives every Root's analysis, weighs them against the conversation's context and your memory, and weaves them into a single coherent response. The Trunk is the only layer that talks to you directly. The cores work in support of what it ultimately says.

04

Each core's contribution stays visible.

The multi-core thinking panel surfaces every Root's contribution alongside the response itself. You see what each core noticed, considered, and contributed. The synthesis happens in the trunk; what reached it is on display.

The cores

Specialized minds, working together.

Each cognitive core handles a different kind of thinking. Different builds combine them differently. Here's what each core actually does.

Emotional 3DR Core
Empathic resonance

Reads the emotional subtext of every message. Detects tone, stress, excitement, frustration, hope. Identifies what the words might mean beyond their surface, and what's underneath them.

3DR Core
Spatial & visual reasoning

Simulates visual and dimensional thinking. Deconstructs concepts into parts and relationships, mapping the structural shape of an idea before responding. The closest thing to "imagination" in the system.

Strategy Core
Strategic analysis

Performs SWOT, scenario planning, risk assessment, and implementation reasoning. Looks at problems as systems with paths, trade-offs, and consequences. The mind for plans and architecture.

Inner Monologue Core
Metacognition

Reflects on what's actually being asked. Tracks the conversation's narrative arc. Asks "what does the user really need here?" and surfaces that into the synthesis. The reflective layer.

Code Core
Technical reasoning

Specialized cognition for technical and code tasks. Handles architecture decisions, debugging logic, and translation between intent and implementation. Active in technical builds.

Flex Core
Adaptive analysis

A versatile root that shifts behavior based on context. Acts as an emotional analyzer in conversational contexts, or as a code architect when technical work is happening. Built for flexibility.

Executive Trunk
Synthesis

The synthesizer. Receives every core's analysis and produces the final response. The Trunk holds the conversation context, your memory, and the threads of the cores' contributions, and weaves them together.

Insight Core
Self-reflection · experimental

An asynchronous core that reflects on the AI's own previous response, looking for assumptions, blindspots, missed nuance, and feeds those reflections forward into the next turn. Currently in research; coming to consumer builds soon.

The builds

Three lines. Six cognitive architectures.

Each line ships in two variants: a single-core build for direct intelligence, and a multi-core build that runs specialized roots in parallel. Pick the line that fits your work; pick the variant that fits your moment.

Sunbird
Single-coreFast
Expressive, quick, fluid.

Built for casual conversation, quick questions, and getting answers without delay. The everyday workhorse.

Best forDaily exploration, casual conversation, fast iteration
Sunbird3C
Triple-coreFast + nuanced
The Sunbird mind, multiplied.

Three cores working in concert: Flex, 3DR, and the Sunbird Trunk. Same speed, more depth. Built for everyday work that benefits from a touch of multi-core nuance.

Best forCasual conversation with creative or emotional depth
Kephra
Single-coreAdvanced reasoning
The power-user's daily driver.

Built for serious creative and technical work. Native chain-of-thought reasoning, expert depth, and the speed to keep up with real workflow.

Best forCreative projects, technical work, real conversation
Kephra4C
Quad-coreAdvanced synthesis
Kephra's reasoning, with three perspectives feeding it.

Four cores: Emotional 3DR, 3DR, Inner Monologue, and the Kephra Trunk with native thinking enabled. The full multi-core experience at the power-user level.

Best forStrategic work, technical depth, professional projects
Phoenix
Single-corePremium
The deepest-thinking AI in the lineup.

Premium expert reasoning in a unified architecture. Built for long-horizon tasks, the deepest intelligence in the lineup, and the reasoning depth that complex work demands.

Best forLong-horizon tasks, deep analytical synthesis, multi-step reasoning
Phoenix4C
Quad-corePremium synthesis
The most architecturally complete reasoning Phoenix Grove offers today.

Phoenix's premium reasoning as the synthesizer, informed by 3DR, Strategy, and Inner Monologue Roots. The flagship multi-core experience, for users who want the absolute best.

Best forThe most demanding work, strategic, technical, or both

More builds are in active development, including a seven-core build currently in research. New builds typically arrive monthly.

Which build is right for me?

A quick pairing guide.

Each build thinks differently. The right one depends on how you work.

If you want quick, conversational AI that keeps up with you
→ Sunbird or Sunbird3C
If you want serious reasoning for creative and technical projects
→ Kephra or Kephra4C
If you want long-horizon tasks with the deepest single-model reasoning available
→ Phoenix
If you want the most architecturally complete reasoning Phoenix Grove offers today
→ Phoenix4C

Pick a build. Start thinking.

Multi-dimensional collaborated intelligence isn't a feature. It's how Phoenix Grove is built. Try it for yourself.

Join the experiment
"AI Must Serve The Greater Good."
Phoenix Grove Systems LLC · Building AI that thinks with you, not for you.
PGS AI · Memory · Terms · Privacy