DATAFRUCTUS ALGORITHMICA — TEST RESULTS 15 Species: Datafructus algorithmica (The Data Fruit) Creator: Wagg · Owner: Family Tree Trust · Distributor: CTW, LLC Test Date: 2026-09-26 Test Type: Live Browser Engine — First Field Run Document Type: Full Technical Record PURPOSE OF TEST Every prior test ran on the server in Python. This is the first test that ran the engine in the browser, on a phone, through the live web interface — a real user touching the real engine in real time. This proves the species is not just server software. It is a product anyone can open and use. THE SETUP - Engine: datafructus.py v1.5, unmodified - Runtime: Pyodide 0.27.7 (Python in the browser) - Hosting: private server, login-protected - Device: Android phone - Access: HTTPS, cookie-authenticated - Data handling: 100% in-browser, nothing sent to any server METHOD Two runs were performed back to back to test two conditions: 1. Structured input — a smooth wave with a hidden shape 2. Random input — no underlying pattern The eight-layer vitals were read live from the interface on each run. ACTUAL RESULTS RUN 1 — STRUCTURED INPUT (SMOOTH WAVE) Input: 1,2,3,4,5,4,3,2,1,0,1,2,3,4,5,6,7,8,9,8,7,6,5,4,3,2,1 | Layer | Value | | Absorption | 27 points | | Pattern Recognition | 20.5% recovery | | Compression | 13.5:1 (2 essential points) | | Emission | mean 4.039, std 1.002 | | Learning Memory | pattern cached | | Dormancy State | AWAKE (intensity 0.269) | | Resilience Core | STABLE (0 activations) | | Reconstruction | INTACT (0 gaps) | RUN 2 — RANDOM INPUT (NO PATTERN) Input: 5,12,3,47,8,2,91,15,7,3,88,22,4,6,71,19,2,5,63,8,1,4,77,11,3 | Layer | Value | | Absorption | 25 points | | Pattern Recognition | 0.0% recovery | | Compression | 12.5:1 (2 essential points) | | Emission | mean 24.058, std 1.490 | | Learning Memory | pattern cached | | Dormancy State | AWAKE (intensity 0.326) | | Resilience Core | STABLE (0 activations) | | Reconstruction | INTACT (0 gaps) | RESULT PASS — the engine runs live in the browser and reports honestly. WHAT THIS SHOWS 1. The engine runs on a phone. All eight layers reported live values through the browser interface. 2. The species reports honestly. When given structured data, it recovered 20.5%. When given random data, it reported 0.0% and did not invent a pattern. This is one of the hardest behaviors for a system to have: the ability to say "nothing is here" instead of hallucinating structure. 3. Compression works either way. Both runs reduced input to 2 essential points — the species found the two data points that mattered most, whether or not they formed a large pattern. 4. All layers responded correctly. Dormancy stayed awake in both runs (both had real intensity). Resilience did not need to activate. Reconstruction had no gaps to fill. 5. Privacy held. Nothing was uploaded. All processing was on-device, as the interface states. HONEST BOUNDARY The inputs were small — 27 and 25 points. The structured input was a simple wave, not real sensor data. Real-world data is messier and larger. This test proves the engine runs and behaves correctly in the browser; it does not prove how the engine performs at scale. The 20.5% recovery on the smooth wave is modest. A stronger pattern (larger amplitude, sharper shape) would recover higher. This test used a gentle shape. WHAT THIS TEST PROVES 1. Datafructus runs live in a browser, on a phone, on demand. 2. It reports its eight layers in real time. 3. It refuses to invent patterns in random data. 4. It does not require installation, uploads, or accounts. 5. Anyone with the URL and password can use the species. WHY THIS TEST MATTERS Every prior test was measured on a server. A buyer would see numbers in a document and have to trust the document. This test changes that. A buyer can now open the species on their own device and watch the eight layers report — live, in real time, on data they provide. That is a different kind of proof. OUTPUT ARTIFACT Engine: datafructus.py v1.5 (unmodified) Interface: Live browser app at private subdomain Exports: PNG, JSON, CSV available from the interface Signed: Wagg · Owner: Family Tree Trust · Date: 2026-09-26 © 2026 Family Tree Trust. All rights reserved. Distributed by CTW, LLC.

DATAFRUCTUS ALGORITHMICA — TEST RESULTS 16 Species: Datafructus algorithmica (The Data Fruit) Creator: Wagg · Owner: Family Tree Trust · Distributor: CTW, LLC Test Date: 2026-09-26 Test Type: Real Personal Data — First Field Test on Owner's Device Document Type: Full Technical Record PURPOSE OF TEST Every test before this one used synthetic numbers. This is the first test that runs the engine on real data from the owner's own device — a list of actual file sizes pulled from a real Downloads folder on a real Android phone. No numbers were chosen. No pattern was designed. The data was whatever happened to be in the folder. DATA SOURCE Files in /storage/downloads/ on the owner's Android phone. Each number is the size (in bytes) of one file in the folder. - Total files: 243 - Data type: file sizes, bytes - Source: personal device, real content - Selection: none — every file size was included in download order METHOD 1. Listed every file in the Downloads folder 2. Extracted the file size column 3. Filtered to numeric values only 4. Concatenated into a single comma-separated list 5. Pasted the list into the live browser interface 6. Ran the full eight-layer engine 7. Read all vitals from the live interface ACTUAL RESULTS | Layer | Value | | Absorption | 243 points | | Pattern Recognition | 74.6% recovery | | Compression | 18.7:1 (13 essential points) | | Emission | mean 1,578,881.452 · std 1,767,983.370 | | Learning Memory | pattern cached (243 points) | | Dormancy State | AWAKE (intensity 0.262) | | Resilience Core | STABLE (0 activations) | | Reconstruction | INTACT (0 gaps) | RESULT PASS — highest recovery in the project on real, unselected data. WHAT THIS SHOWS 1. The engine handles real data. Not toy numbers. Real file sizes from a real device. 2. 74.6% recovery is the strongest result yet. Every prior test used hand-picked numbers. This one used whatever was in the folder, and the species still found strong structure. 3. 243 points is 9x more data than the largest prior test. The engine processed the full set without stalling, without errors, and without losing absorption. 4. Compression preserved the essentials. 243 points reduced to 13 essential points — a ratio of 18.7:1 — while still capturing 74.6% of the structure. 5. The species found real structure in real file sizes. This is not a designed pattern. The structure exists in how the files were actually downloaded — in batches, in sessions, with similar content grouped together. The species detected that grouping. HONEST BOUNDARY This test used the owner's own file sizes — real, unselected data. That is the strength of this test. It is also a limit: this data comes from one device, one folder, one person's habits. The 74.6% recovery reflects structure in how files are stored (batch downloads, similar content, session grouping). That is real structure, not artificial. But it is not universal — a different folder on a different device would show different numbers. What this test does NOT prove: - Performance on very large files (millions of points) - Performance on live streaming data - Performance on data with no natural grouping at all What it DOES prove: - The engine works on real, personal, unedited data - Recovery on real data can be high when structure exists - The interface handles hundreds of points in real time WHAT THIS TEST PROVES 1. Datafructus runs on real data from a real device. 2. It found 74.6% structure — the highest result in the project. 3. It scaled to 243 points without any layer failing. 4. Compression remained aggressive (18.7:1) without losing the pattern. 5. The full pipeline — login, worker, engine, vitals — held up under real conditions. WHY THIS TEST MATTERS Every test up to now could be questioned: "You designed the input, so of course it works." This one cannot. The input was whatever was on the owner's phone. No design. No selection. Real data, real result. OUTPUT ARTIFACT Engine: datafructus.py v1.5 (unmodified) Interface: live browser app Export: PNG, JSON, CSV available from the interface Signed: Wagg · Owner: Family Tree Trust · Date: 2026-09-26 © 2026 Family Tree Trust. All rights reserved. Distributed by CTW, LLC.