AtMem's memory classifier rejected all 100 items in a replay
In an October 17 community research post, Javad Taghia reported why AtMem did not ship a fine-tuned memory classifier. Its fully correct test transitions rose from 18.4% to 73.8%. A rule that always rejected storage scored 91.1% on the same data. A later version rejected all 100 replay observations.
Verified 12:56 AM PDT · 1 original sources
The developer ran this small experiment, not an independent lab. The replay used fictional items that the developer had inspected during development. The deterministic baseline made only 30 correct decisions out of 100.
Independent replays should include rare cases when the system needs to save a memory. Tests should count missed memories and later corrections, not only overall accuracy.
Audit the story
Original sources
Company claims remain company claims. Follow the reporting and judge the evidence directly.
Continue the edition