Open the
black box.
A place to see how the pieces fit.
Not just what comes out at the end.
These are plans for future interactive experiments. The blueprints below explain the direction; they do not run live models or produce benchmark results.
A question, with a visible path.
A future walkthrough of how a question finds relevant information, how that information becomes context, and where the final answer gets its sources.
What the future version will let you do
- Inspect the original question and retrieved passages.
- See what is kept, what is excluded, and why the context matters.
- Connect statements in the answer back to their sources.
See the steps between ask and answer.
A future view into an agent workflow: the plan, the tools it calls, the state it carries, and the checks before the next action.
What the future version will let you do
- Follow a plan through a sequence of tool calls.
- Inspect inputs, outputs, and the state between steps.
- See how failed steps and human approval are handled.
Different constraints. Different choices.
A future comparison space for model quality, latency, and memory. Real measurements will be added with their hardware, task, and evaluation setup.
What the future version will let you do
- Compare runs made under the same test conditions.
- See hardware, runtime, quantization, and context settings.
- Look beyond one score to understand the actual trade-offs.
Complexity,
from another angle.
A small interactive systems sculpture. Move your pointer
over it or use the arrow keys to change your point of view.
A conceptual visualization, not a model or a live simulation.