SRT Omni Cross-Vendor Retrieval
🔀
Search mixed-media index with text across AI models
SRT demos
Read-outs on frozen models you can actually run. Nothing here fine-tunes the host.
Note Four companies, one searchable memory. The newest and the strongest.
ChestX-ray14 frozen probe
🫁1339 KB linear probe beats a fine-tuned ResNet-50
Note A 339 KB linear probe on frozen features reads chest radiographs, ahead of the fine-tuned baseline on the official split.
0.6B Reads 27B
🪟A 0.6B model in your tab searches a 27B model's index
Note A 0.6B model puts words to a 31B's raw internal state, scored against an index it did not build.
One Model, Two Jobs
🪟A 0.6B model in your tab searches a 27B model's index
Note The read-out running on the visitor's own hardware.
SRT introspect
🧭4Adaptive-density reasoning traces over a frozen Qwen-2.5-7B
Note Read the state of a frozen model while it runs.
Gemma-4-31B-it SRT-Sunstone
🔮3A read-out that reads images — trained only on words
Note Gemma-4 31B with an SRT read-out attached.
SRT Showcase
🧭2Watch a frozen Qwen-2.5-7B think — live SRT introspection
Note The SRT program end to end.
MindReader-NLA
🧠3Ask a frozen LM what it is thinking, in plain English.
Note Natural language access to a frozen model's activations.
SRT-Adapter v8a Demo
🌊3Per-token reflexivity heatmap from a frozen Qwen2.5-7B
Note The adapter that started the line.