Global Intent Company · Independent AI & systems engineering

Try the tools.
Inspect the research.
Build on the evidence.

Chat and research with LooseMouth, test GIC's experimental language models, or run a learning-system experiment in your browser.

LooseMouth uses Google Gemini. GIC's own model checkpoints run separately on Hugging Face. Private model infrastructure is an ongoing research program.

Create free account Compare plans

Try chat before signing up. Create an account to save conversations.

Developer quickstart and reproducible examples

Start with a question

5 guest messages. No account or payment card needed to try.

Your draft stays in this browser tab for up to 10 minutes. Review it in the workspace before sending.

Run BitVision Cell Free · no login

Download the checkpoint, then run inference locally with WebGPU or WASM. Inspect actions and episode trajectories.

Explore the Virtual Lab Paid / Enhanced

Review assumed kinetics, generate synthetic optics, and export evidence. An active subscription is required.

Choose your next step

Working tools, with their scope in view.

Start with a free experiment, inspect the implementation, or open a subscription workspace when you need more.

LooseMouth · Google Gemini

Chat, research, and reusable reports

Guest chat available

Start with a question. Create a free account to save conversations; Paid adds reports and Enhanced adds interactive applets.

Try LooseMouth

INTENT RLSE-v2 · 187.3M parameters

Test a model trained by GIC

Public research demo

Compare Facts-500 and SFT-1000 with streaming output and adjustable decoding. These experimental checkpoints are not production assistants.

Test the checkpoints

BitVision Cell · v0.2

Inspect a learning system in motion

Free · browser-local inference

Load an ONNX checkpoint and inspect its actions and episode trajectory. Runtime and model files download before computation runs in your browser.

Run BitVision Cell

Virtual Lab · research cockpit

Explore a scientific workflow

Paid / Enhanced subscription

Approve assumption-based kinetics, inspect reference structures, generate synthetic frames, and check exported hashes. No clinical prediction or hardware measurement is implied.

Open the paid Virtual Lab

Developer quickstart

Reproduce a request

Public documentation

Copy working streaming examples, inspect limits and output formats, and follow implementation notes with source links and known limitations.

Read the developer quickstart

Research & private infrastructure

Inspect the architecture

Source, experiments, and design goals

Read the evidence behind INTENT and the PLMH / PLMN infrastructure program. Research goals are distinguished from deployed capabilities.

Inspect research and evidence

Keep the work you start.

Create a free account to save LooseMouth conversations. Compare Paid and Enhanced for reports, applets, and Virtual Lab access.

Create free account Compare access plans

Try the tools.Inspect the research.Build on the evidence.