Type a section name and press Enter. Use “help” for the list.
Figures: softened for dark viewingDisplay only; image files are unchanged.
Cliste experiment
1.2 — photographs, video and pauses in the same life
Registered 7 Oct 2026 · Snapshot 12 Oct 2026
Execution: completedScientific verdict: pass
The question
Continue the existing VSA life across uncaptioned photo folders, streamed real video, pauses and graceful interruption.
Why this experiment?
Before testing reasoning or understanding, a continuing system needs to admit experience, retain it and recover it after a restart.
How it works
Admit natural images into VSA memory, add more experience, then decode earlier vectors and test restoration from a saved state. The earliest image interface used small center crops, so it did not retain the complete scene.
Follow this part of the investigation: Can an experience survive after its image file is gone? →Implementation and comparison details
Extends the twelve retained 1.1 experiences with caption-free folder intake, bounded video decoding, canonical tensor batches and paired state/cursor checkpoints.
What happened
All registered mixed-media, pause, graceful interruption, source-free readback and continuation checks passed.
All registered mixed-media, pause, graceful interruption, source-free readback and continuation checks passed.
Measurements
| Recorded field | Value |
|---|---|
| episodes | 72 |
| inherited | 12 |
| new photos | 12 |
| new video frames | 48 |
| mean mse all | 0.004184882875405388 |
| mean mse inherited | 0.00645667602327277 |
| mean mse new photos | 0.006033867374366665 |
| mean mse andor | 0.0024742599088117554 |
| mean mse dune | 0.003835117018584692 |
| restart max abs | 0 |
| paused restart max abs | 0 |
| inherited readback max abs | 0 |
| resources | {
"peak_rss_bytes": 1584807936,
"peak_cuda_allocated_bytes": 285352960,
"peak_cuda_reserved_bytes": 322961408
} |
| final storage | {
"logical_content_bytes": 18874368,
"logical_context_bytes": 1179648,
"allocated_content_bytes": 33554432,
"allocated_context_bytes": 2097152,
"peak_rss_bytes": 1584807936,
"peak_cuda_allocated_bytes": 285352960,
"peak_cuda_reserved_bytes": 322961408
} |
| snapshot bytes | 43333209 |
| intake receipts | [
{
"status": "paused",
"episodes": 20,
"admitted": 8,
"batches": 2,
"next_ordinal": 8,
"snapshot": "20261007T183603185791Z-000014",
"metrics": {
"seconds": 0.6846521330298856,
"encode_seconds": 0.5808024989673868,
"checkpoint_seconds": 0.06982518103905022,
"checkpoints": 1,
"max_buffered_images": 4,
"decoded_frames": 0,
"emitted_samples": 8,
"max_preprocessed_in_reader": 1,
"effective_batch_size": 4
}
},
{
"status": "complete",
"episodes": 24,
"admitted": 4,
"batches": 1,
"next_ordinal": 12,
"snapshot": "20261007T183616556651Z-000017",
"metrics": {
"seconds": 0.5773473340086639,
"encode_seconds": 0.4868350039469078,
"checkpoint_seconds": 0.07654814596753567,
"checkpoints": 1,
"max_buffered_images": 4,
"decoded_frames": 0,
"emitted_samples": 4,
"max_preprocessed_in_reader": 1,
"effective_batch_size": 4
}
},
{
"status": "paused",
"episodes": 32,
"admitted": 8,
"batches": 2,
"next_ordinal": 432,
"snapshot": "20261007T183620945858Z-000018",
"metrics": {
"seconds": 1.1944197789998725,
"encode_seconds": 0.54062146297656,
"checkpoint_seconds": 0.07958611904177815,
"checkpoints": 1,
"max_buffered_images": 4,
"decoded_frames": 409,
"emitted_samples": 8,
"max_preprocessed_in_reader": 1,
"effective_batch_size": 4
}
},
{
"status": "complete",
"episodes": 48,
"admitted": 16,
"batches": 4,
"next_ordinal": 816,
"snapshot": "20261007T183626598177Z-000019",
"metrics": {
"seconds": 1.8287647460820151,
"encode_seconds": 0.5572516059037298,
"checkpoint_seconds": 0.10757457907311618,
"checkpoints": 1,
"max_buffered_images": 4,
"decoded_frames": 816,
"emitted_samples": 16,
"max_preprocessed_in_reader": 1,
"effective_batch_size": 4
}
},
{
"status": "complete",
"episodes": 72,
"admitted": 24,
"batches": 6,
"next_ordinal": 816,
"snapshot": "20261007T183632741831Z-000020",
"metrics": {
"seconds": 1.5960504428949207,
"encode_seconds": 0.5182606258895248,
"checkpoint_seconds": 0.0999684960115701,
"checkpoints": 1,
"max_buffered_images": 4,
"decoded_frames": 816,
"emitted_samples": 24,
"max_preprocessed_in_reader": 1,
"effective_batch_size": 4
}
}
] |
| max buffered images | 4 |
Recorded checks
| Recorded field | Value |
|---|---|
| tamper probe home unchanged | Pass |
| distinct host process launches | Pass |
| observer existed before removal | Pass |
| original 1 1 home unchanged | Pass |
| final isolation home unchanged | Pass |
| actual sigint acknowledged | Pass |
| observer absent before resume | Pass |
| paused isolation home unchanged | Pass |
| photos first:exact save load cursor | Pass |
| photos first:frozen model | Pass |
| photos first:recall and render state contract | Pass |
| photos first:restored model exact | Pass |
| photos first:nonworking unchanged | Pass |
| photos first:earlier rows unchanged | Pass |
| photos first:exact save load state | Pass |
| tamper:changed source cursor unchanged | Pass |
| tamper:changed source refused | Pass |
| tamper:changed source state unchanged | Pass |
| paused:exact save load cursor | Pass |
| paused:pause preserves cursor | Pass |
| paused:exact save load state | Pass |
| paused:frozen model | Pass |
| paused:pause preserves episode banks and working | Pass |
| paused:pause advances context | Pass |
| paused:recall and render state contract | Pass |
| paused:pause recall only changes working | Pass |
| paused:restored model exact | Pass |
| paused:nonworking unchanged | Pass |
| paused:earlier rows unchanged | Pass |
| paused:pause render changes nothing | Pass |
| isolated paused:frozen model | Pass |
| isolated paused:isolated pause render changes nothing | Pass |
| isolated paused:observer inaccessible | Pass |
| isolated paused:readback preserves loaded state | Pass |
| isolated paused:recall and render state contract | Pass |
| isolated paused:nonworking unchanged | Pass |
| isolated paused:isolated pause advances context | Pass |
| isolated paused:isolated pause preserves banks and working | Pass |
| isolated paused:sources inaccessible | Pass |
| isolated paused:all source probes denied | Pass |
| photos resume:exact save load cursor | Pass |
| photos resume:frozen model | Pass |
| photos resume:recall and render state contract | Pass |
| photos resume:restored model exact | Pass |
| photos resume:nonworking unchanged | Pass |
| photos resume:earlier rows unchanged | Pass |
| photos resume:exact save load state | Pass |
| andor interrupt:exact save load cursor | Pass |
| andor interrupt:frozen model | Pass |
| andor interrupt:recall and render state contract | Pass |
| andor interrupt:restored model exact | Pass |
| andor interrupt:graceful signal stopped at two batches | Pass |
| andor interrupt:nonworking unchanged | Pass |
| andor interrupt:earlier rows unchanged | Pass |
| andor interrupt:exact save load state | Pass |
| andor resume:exact save load cursor | Pass |
| andor resume:frozen model | Pass |
| andor resume:recall and render state contract | Pass |
| andor resume:restored model exact | Pass |
| andor resume:nonworking unchanged | Pass |
| andor resume:earlier rows unchanged | Pass |
| andor resume:exact save load state | Pass |
| dune:exact save load cursor | Pass |
| dune:frozen model | Pass |
| dune:recall and render state contract | Pass |
| dune:restored model exact | Pass |
| dune:nonworking unchanged | Pass |
| dune:earlier rows unchanged | Pass |
| dune:exact save load state | Pass |
| isolated final:frozen model | Pass |
| isolated final:isolated pause render changes nothing | Pass |
| isolated final:observer inaccessible | Pass |
| isolated final:readback preserves loaded state | Pass |
| isolated final:recall and render state contract | Pass |
| isolated final:nonworking unchanged | Pass |
| isolated final:isolated pause advances context | Pass |
| isolated final:isolated pause preserves banks and working | Pass |
| isolated final:sources inaccessible | Pass |
| isolated final:all source probes denied | Pass |
| uninterrupted:exact save load cursor | Pass |
| indexOmission | 32 further fields |
Results and visual artifacts
Open a figure for its input, comparison, and result. Click an image to enlarge it.
Each group compares the exact 64 × 64 input supplied to Cliste with its reconstruction from stored vectors, then the reconstruction after restarting without access to the original images. Observations 0–11 are inherited photographs; 12–23 are new photographs; 24–47 and 48–71 are sampled Andor and Dune trailer frames. These are stored-image readbacks by observation number, not a test of recognizing an image from a partial cue. Video frame numbers identify the sampled source frames; PTS is the video’s presentation timestamp, not the system’s internal encounter count.
Inherited photographs: reconstruction before and after restart, observations 0–5
Replotted from saved outputs with clarified labels. The original recorded figure remains available.
Inherited photographs: reconstruction before and after restart, observations 6–11
Replotted from saved outputs with clarified labels. The original recorded figure remains available.
New photographs: reconstruction before and after restart, observations 12–17
Replotted from saved outputs with clarified labels. The original recorded figure remains available.
New photographs: reconstruction before and after restart, observations 18–23
Replotted from saved outputs with clarified labels. The original recorded figure remains available.
Andor video frames: reconstruction before and after restart, observations 24–29
Replotted from saved outputs with clarified labels. The original recorded figure remains available.
Andor video frames: reconstruction before and after restart, observations 30–35
Replotted from saved outputs with clarified labels. The original recorded figure remains available.
Andor video frames: reconstruction before and after restart, observations 36–41
Replotted from saved outputs with clarified labels. The original recorded figure remains available.
Andor video frames: reconstruction before and after restart, observations 42–47
Replotted from saved outputs with clarified labels. The original recorded figure remains available.
Dune video frames: reconstruction before and after restart, observations 48–53
Replotted from saved outputs with clarified labels. The original recorded figure remains available.
Dune video frames: reconstruction before and after restart, observations 54–59
Replotted from saved outputs with clarified labels. The original recorded figure remains available.
Dune video frames: reconstruction before and after restart, observations 60–65
Replotted from saved outputs with clarified labels. The original recorded figure remains available.
Dune video frames: reconstruction before and after restart, observations 66–71
Replotted from saved outputs with clarified labels. The original recorded figure remains available.
Limits and next question
One frozen 64×64 sensory candidate, twelve added COCO validation photographs and two sampled trailers. Prior model evaluation used the validation split. Reconstruction concerns admitted crops; frame selection omits most video. Explicit episode readback does not establish partial-cue recall, cleanup, learning or understanding.
Next question. Review bounded intake, interruption continuity and visible fidelity before proceeding to partial-cue retrieval in 1.3.
Inspect the evidence
Download the methods and measurements for this experiment. Large training and saved-state files are not included.