AI-generated Eloquent relationship loop causing massive N+1 performance degradation and unindexed foreign key lookups.
$posts = Post::all();
foreach ($posts as $post) {
echo $post->user->name;
}Demonstrate deep domain expertise against superficial AI prompting.
"I know database design and Laravel inside and out. Calling $post->user inside a loop without eager loading fires N+1 separate queries. I'd replace this with Post::with('user')->get() and ensure foreign keys have composite indexes."
The trainer shows a posts loop accessing a related user on each iteration. Under the illustrated uncached lazy-loading assumption, one initial query plus a separate relationship query for each of one hundred posts gives one hundred one roundtrips. Eager loading the related users in a separate batch gives two. Bars encode these hypothetical counts at three pixels per query. Actual database behavior depends on relationship caching, constraints, selected columns and data, and this page runs no PHP, database connection or SQL benchmark. Its displayed remediation is a code example and selected explanation, not a measured reduction applied to an actual application. The mechanism matters more than accepting a fixed mastery label. The default scenario offers three defense choices, each carrying authored confidence, rating, statement and query caption. Bars encode three available choices and one selected choice at sixty pixels per item. Choosing the shallow response changes confidence to Low and rating to Vulnerable to Hallucinations, but the export still includes the same eager-loading remediation diff from the scenario table. Thus it records a selected response and a supplied answer rather than a patch proven by the user. Native option-card click and actual JSON download verify those changes in independent original and preview contexts. A categorical confidence label is not an independently scored demonstration of competence. The source includes three predefined scenarios covering relationship loading, concurrent balance updates and missing indexes. Bars compare the count three with one constant boolean flag, explicitly different meanings. The export assigns proves-human-oversight true for every response, including the shallow choice. Mastery-level instead checks whether the selected confidence string equals High. Neither predicate tests reasoning, executes a transaction race, checks an index plan or verifies an actual candidate identity. The trainer is a practice rubric with supplied statements. Original interactive cards, export panel, static SVG and all bottom footers retain their dimensions through four native viewport transitions. No real application repair, public publication or exhaustive skill certification is claimed.