Memoria Vivum — Official Test Report

Test Objective

To test the core living archive behavior in real, runnable Python code: absorption, connection, recall, forgetting, and verification across five growth rounds. This is the first Memoria vivum test that anyone can run and reproduce.

Test Method

A knowledge base of 15 items was created: 10 meaningful facts and 5 noise items. Each item carries tags. The species processed the base over 5 rounds, measuring:

• Absorption: items taken into memory
• Connections: relationships formed by shared tags
• Recall: true facts retrievable with understanding
• Noise Forgotten: irrelevant items released
• Verification: true items confirmed by connection depth

Actual Results

Final

• Truth Retained: 10/10
• Noise Forgotten: 5/5
• Verified True: 3/10
• Total Connections: 10
• Items in Memory: 10

Result

PARTIAL PASS

Interpretation

What Worked

• Truth retained: 100%. All 10 true facts survived all five rounds.
• Noise forgotten: 100%. All 5 noise items were released by the end of the run.
• The living archive is real. Memory grew, connections formed, noise decayed, truth held.

What Struggled

• Verification reached only 3/10. Only three facts built enough connections in this run.
• Total connections low (10). The connection engine requires shared tags, and the current tag sets are too sparse.

The Core Finding

The species protects what matters (truth) and releases what does not (noise). But it does not yet build a rich enough web of connections to verify most of what it holds. This tells the species exactly what to evolve next.

Lessons Learned

1. Forgetting and truth-protection work perfectly out of the box.
2. Verification depends on connection depth, which depends on how richly items overlap.
3. The next evolution must deepen the connection engine.

Next Steps

• Test MV-006: deepen the connection engine so verification climbs higher.
• Keep the honest measure: final state, not cumulative count.

Output Artifact

Results file: test_mv05_baseline.txt
Script: scripts/test_mv05_baseline.py

Memoria Vivum — Official Test Report

Test Objective

To test the v1.6 evolution. The Weaving Core and the Verification Core were both extended. This test measures whether the extensions improved verification compared to v1.5 (MV-005, which reached 3/10 verified).

What Was Extended in v1.6

Weaving Core — was extended to find connections that run through other items, not only direct ones. The original direct matching still works exactly as designed.

Verification Core — was extended to confirm a fact when it is supported by two or more connection paths (direct plus extended). The original verification logic still works exactly as designed.

Test Method

Same knowledge base as MV-005: 10 true facts, 5 noise items. Five rounds. Measured absorption, connections, recall, noise forgotten, and verification.

Actual Results

Final

• Truth Retained: 10/10
• Noise Forgotten: 5/5
• Verified True: 5/10
• Total Direct Connections: 10
• Total Extended Connections: 6
• Total Connections: 16

Result

PARTIAL PASS — Evolution confirmed, target not reached.

Comparison to v1.5

Verification rose from 3/10 to 5/10. A 67% improvement on the same knowledge base.

What This Proves

1. The evolution worked. The extended Weaving Core found 6 connections that v1.5 could not see.
2. Truth retention and noise forgetting stayed perfect. The extensions did not damage any existing capability.
3. The Verification Core now confirms more facts because it has more connection paths to work with.

Honest Boundary

Verification reached 5/10, not the predicted 8/10. The knowledge base is sparse — 10 facts, each with 3 tags. Most facts still have only 1-2 connection paths. The verification rule requires 2 or more paths. Half the facts sit one short.

This is a property of the knowledge base, not a failure of the species. A richer knowledge base with more overlapping items would give the Verification Core more to work with.

Lessons Learned

1. The extended Weaving Core works. It found real connections that direct matching missed.
2. The Verification Core works. It confirms more when given more paths.
3. Connection density depends on the knowledge base itself. The species cannot verify what it cannot connect.

Next Steps

• Test on a richer, real knowledge base to measure verification at realistic density.
• Document honestly whether verification climbs toward the 8/10 target under real conditions.

Output Artifact

Results file: test_mv06_evolved.txt
Script: scripts/test_mv06_evolved.py