Media + independent replication
What can be said, and what still needs proving.
One-sentence description
Fractapp FIC/5 is an experimental lossless image-storage architecture that gives image regions recursive mathematical addresses, lets each region choose a reversible local encoding, and retrieves only the blocks needed for a requested crop.
Maximum supported result statement
On the declared public split and subsequently opened H2 holdout, the frozen FIC/5 balanced profile was smaller in total, moved fewer random-access crop bytes, and decoded faster at the median than the strongest exact-deduplicated complete 64/128/256 tile-pyramid baselines implemented in the experiment, while reconstructing every tested crop exactly.
Do not say
- “FIC/5 beats PNG/WebP/AVIF/JPEG XL.” That has not been established.
- “Fractals magically compress arbitrary data.” They do not.
- “The result is independently replicated.” It is not yet.
- “The H2 holdout is reusable.” It is consumed.
Attribution
Jonathan H. Lindley and ChatGPT, 7 August 2026. Jonathan H. Lindley originated and directed the research program. ChatGPT contributed algorithm design, engineering, benchmark design, analysis, and drafting as an AI co-author. This does not imply OpenAI review or endorsement.
Replication checklist
- Use the frozen FIC/5 balanced implementation.
- Keep the declared baseline matrix unchanged.
- Verify exact source-crop hashes.
- Report total, payload, cold-object, and decode ratios separately.
- Disclose hardware/runtime and all deviations.
- Use a fresh holdout generation for any new optimization cycle.