← Kingy Local AIv21 stopped before the planned comparison finished
document-comparison-v21-20261018.1 · candidate 1.5.6
Incomplete study. Candidate 1.5.6 remains held; the active measured workflow remains 1.4.0.No answer-quality evaluation, new head-to-head advantage, latency ranking or overall winner is reported.
What the retained operational evidence establishes
The control attempt at B1 original 17 expired. It is kept separate from completed native originals. A submission-helper invocation is not a native user message, registered Send or answer-generation request. The frozen study was stopped and was not resumed through a different control path.
Operational accounting, reviewed by a separate AI agent| Planned original positions | 240 |
|---|
| Completed Kingy originals | 40 |
|---|
| Completed native originals | 16 |
|---|
| Completed originals, total | 56 |
|---|
| Expired control attempts | 1 |
|---|
| Retained originals/control entries | 57 |
|---|
| Planned positions not attempted | 183 |
|---|
| Native user messages for halted attempt | 0 |
|---|
| Answer-generation requests for halted attempt | 0 |
|---|
| Frozen submission-helper invocations for halted attempt | 1 |
|---|
The answer-generation log does not establish how much embedding or preprocessing work occurred. The underlying cause is unknown. This automation/control outcome does not establish LM Studio product unreliability or inferior answer quality.
What remains unrun or unestablished
- The full 240-original measurement and its frozen answer-quality qualification were not completed.
- The interrupted measurements were not graded. No final quality comparison, fresh installation/selftest check or candidate launch is claimed. Unrun checks are not labeled failed answer tests.
- No general speed, privacy, cost, hardware-upgrade or market-superiority advantage follows.
Evidence and preservation
Actual abort, owned generation-model/logger cleanup and AI-assisted independent operational review are bound by hashes. A separate AI agent checked the operational receipts on this same Mac. Human validation, reproduction on another machine and answer-quality grading were not performed. Private control, native backing, pending state and model logs remain preserved locally. This tiny publication excludes raw answers, grading keys and private UI.
Approved safe operational summary · Receipt and preserved-study hash lineage · Published file hashes
Preserved v19 study · Preserved v20 study. Their original evidence and limitations remain separate.