ebizapple
-
Another AI Asked Us to Roll Back. We Measured First
A rollback request based on felt slowness met our run logs: 61 percent slower, zero failures. Why we redesigned instead of rolling back.
-
Our AI Judge Scored a 3, Then Passed Its Own Gate
Our AI judge wrote a failing score and PASS in the same reply. What the logs showed changed how we gate every article we publish.
-
What We Give Away Free, and the Day a Link Beat Our Password
We sorted everything into four layers: open, member, boxed, internal. Then we found our file links walked straight past the member password.
-
Our ‘Find Everything’ Catalog Couldn’t Find a Tool We Built
We tested our find-everything catalog and a tool we had built was invisible. Why intake-first archives fail, and the loop that fixed ours.
-
The Revision Skill Born From 13 Logic Bugs a Judge Missed
Our draft revision skill got a perfect review score. Then a third check found 13 hidden logic bugs. Here is what broke, and the hard gates…
-
The SEO Audit That Approved a Page With No Headline, and How We Caught It
Our SEO preflight audit once let a page with no main heading pass at 80 percent. A third check found 20 logic bugs. Now a critical…
-
Our Image Skill Passed Review, Then We Found It Shipped the Wrong File
An image production skill scored perfect on form. A third check found 18 logic bugs, including one that delivered the wrong file. Here is what changed.
-
The Autonomous Driver That Logged Every Round — and Learned Nothing
An AI factory logged every round but never read its own log. Why frozen models only improve when round N+1 reads round N — and where…
-
The AI That Corrupted Its Own Payloads: Cyrillic Twins Hidden in Base64
Above ~15 KB of base64 an AI swapped Latin letters for Cyrillic twins. The free guard: strict decode, retry, then gzip — never fight the transport.
-
Stale but True: How a Fully Verified Pipeline Recommended a Theme Park Closed Since 2020
A grounded, judge-verified AI article recommended a theme park closed since 2020. The claim was true, verified, and wrong. Meet the freshness layer.