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.
Watch the cores think in parallel.
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.
Four streams of thought, braiding into one response.
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.
How a multi-core build actually responds.
The path your message takes from "send" to "answer," in four steps.
Four Roots thinking in parallel. One Trunk speaking with one voice.
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.
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.
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.
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.
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.
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.
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.
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.
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.
Specialized cognition for technical and code tasks. Handles architecture decisions, debugging logic, and translation between intent and implementation. Active in technical builds.
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.
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.
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.
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.
The everyday line: built for conversation that keeps pace with you, and answers that arrive without delay.
Casual conversation, quick questions, fast iteration. The everyday workhorse.
Three cores working in concert: Flex, 3DR, and the Sunbird Trunk. Same speed, more depth.
Serious reasoning with real speed, for creative and technical work that has substance.
Native chain-of-thought reasoning, expert depth, and the speed to keep up with real workflow.
Four cores: Emotional 3DR, 3DR, Inner Monologue, and the Kephra Trunk with native thinking enabled.
Premium reasoning for long-horizon work and the questions that demand everything.
Premium expert reasoning in a unified architecture, with the depth that complex, multi-step work demands.
Phoenix's premium reasoning as the synthesizer, informed by 3DR, Strategy, and Inner Monologue Roots.
More builds are in active development, including a seven-core build currently in research. New builds typically arrive monthly.
A quick pairing guide.
Each build thinks differently. The right one depends on how you work.
Pick a build. Start thinking.
Multi-dimensional collaborated intelligence isn't a feature. It's how Phoenix Grove is built. Try it for yourself.