Ternary Computing (TCN)
{-1, 0, +1}
True low-bit inference for
extreme efficiency.
Internectual Labs — Efficient AGI & Experimental Architectures
The Internectual Philosophy
“INTERNECTUAL (adj): Someone who has cultivated a substantial body of knowledge via the web.”
Today's minds are always on the internet, but it's what you do with it that sets you apart. — Cody Dixon.
Engineering Pillars
Grounded Agentic Logic
Safe, secure, and logically
grounded decision making.
Experimental AGI Systems
Novel transformer variants and
reasoning loops.
Inside the lab
Ideas worth building.
Independent research into efficient AI, new model architectures, and tools that make complex systems easier to work with.
Wave models. A different direction.
Exploring AI models that use I/Q signals—in-phase and quadrature components—to work with wave-based representations. This is an experimental research direction, with the focus on what these representations could offer learning and reasoning.
Research in progressMake smaller models accomplish more.
Ternary computing, low-bit models, and post-quantization training are central to the lab's work. The aim is useful reasoning with a smaller footprint, measured against the trade-offs in quality, memory, and speed.
Model developmentSee the system you're building.
AI workflows should be easier to inspect. From model maps to connected flow blocks, the lab is developing ways to see how components fit together and follow the path through a complex system.
Tools in developmentKeep the goal in view.
Building tools for AI-assisted projects: keeping track of progress, coordinating work, and checking what actually changed. The focus is practical follow-through, with the user choosing the objective and limits.
Agent toolingProject directory
Built here. Taking shape here.
A look at the models and tools underway at Internectual Labs. Projects marked Not released are still in development.
5 projects shown
MODEL / TERNARYModel page
TAARDIS-27B
The lab's full-ternary model project. Explore the model page for its published files, documentation, and usage information.
Explore on Hugging FaceTOOL / AGENT WORKFLOWGitHub repository
GRIP
Tell your AI to get a GRIP. On-demand project continuation, saved progress, and a record of what changed and what was tested. You choose when it runs.
Explore GRIP on GitHubDESKTOP / VISUAL AINot released
GNU-AI-Flow
A fork of GNU Radio Companion that brings its desktop flow-block interface to AI work. Connect and inspect AI components, model flows, and training workflows through a visual graph.
Built around the GNU Radio Companion interface, as a separate AI-focused application.
In development · Release planned when readyMODEL / SIGNAL RESEARCHNot released
IQ Wave Models
Experimental AI architectures using in-phase and quadrature signals. Exploring wave-based representations as another way to approach learning, information, and reasoning.
Experimental research · No release date announcedTOOL / MODEL EXPLORATIONNot released
Qwen Atlas
An interactive view of a model's layers and blocks. Designed to make architecture, precision choices, and model-size trade-offs easier to explore visually.
In development · No release date announced