INTENT
Compact experimental language-model architectures exploring parameter efficiency, training behavior, and reasoning under constrained compute.
Model research
Follow the work on recurrent language models, training, inference, and systems designed to be controlled by their operators.
INTENT model family
INTENT is GIC’s model-development program, spanning compact experimental architectures and the public LooseMouth research interface.
Compact experimental language-model architectures exploring parameter efficiency, training behavior, and reasoning under constrained compute.
GIC’s recurrent language-model research direction, extending the INTENT model-development program.
The public workspace for chat, web research, reports, and applets. Hosted inference currently uses Groq while GIC develops its private models and infrastructure.
BitVision has its own research overview and live browser-local Cell demonstration.