An answer is only the surface.
Explore the computation underneath.
ATLAS captures model execution. Shadow State is our next step: a one-way window into the running model.
Research foundation. Runtime layer in development.ATLAS gives captured model execution a structure. Tokens, layers, routes and retained internal states become connected observations.
Meet the capture platformOne index connects the record: a run, a token, a layer, an operator. Capture coverage and retained evidence stay part of the story.
Navigate an illustrative traceShadow State is being designed to stream selected internal observations to an external observer. A one-way path out. No shadow memory fed back in.
Explore the runtime directionDifferent observations answer different questions. Explore what the research capture pipeline records—and where the boundaries remain.
Follow captured prompt and generated-token sequences through indexed model execution. Keep positions, model identity and capture context attached to the observations.
Generated text is an output record, not guaranteed access to a model’s private reasoning. Historical short-prompt runs were capped at 32 generated tokens.
Move through an example execution. Select a token and layer, switch the signal, and follow the same coordinate across the view.
This explorer demonstrates navigation, not the private ATLAS schema. Values are generated locally; they are not measurements, safety scores or model internals.
The runtime direction: selected model observations leave through an indexed stream. Storage and analysis stay outside the model’s accessible context.
Normal model-owned
execution and state.
Selected observations.
External, indexed stream.
Archive, inspect and
evaluate monitoring.
One-way by design. No Shadow State fed into model attention. Non-interference, capture overhead and isolation need validation in each runtime.
A compact, continuous view of selected observations. Intended for baseline comparisons and escalation to richer capture.
DESIGN DIRECTION · NOT A SHIPPING TIERATLAS has an internal model-capture research foundation. Shadow State extends that work toward runtime observability. The next step is validation—not a stronger promise.
From Python to GPU kernels.
In the cross-domain capture series.
In the instrumented K3 runtime.
Founder-supplied research records. Short-prompt runs with a 32-generated-token cap; not 252 complete agent tasks, an independent audit or a security benchmark. Read the scope ↗
Configured forward capture, routing trajectories, selected memory experiments and objective-linked backward analysis.
Inside ATLASModel-wide Shadow State extraction, bounded streaming and complete live agent lifecycle integration.
View development boundariesOutlier detection, pre-action warning, intervention and alignment feedback remain research questions.
Read the open questionsStart with a model, a question, and the evidence
you would need to trust the result.