Many have experienced the effects from AI slop and workslop. I thought it’s time to coin a new term to describe it. Ironically defined by an AI generated dictionary entry:

shit left | noun | /ʃɪt left/

Definition: The contamination of a product development pipeline through the introduction of low-quality, AI-generated content (“AI slop”) at the earliest stages, resulting in compounding degradation of information quality throughout all subsequent phases of production.

Etymology: Portmanteau of “shift left” (the software engineering practice of moving testing and quality assurance earlier in the development cycle) and vulgar slang, ironically inverting the positive connotation of early-stage intervention.

Usage: Technical, informal

The phenomenon follows the principle of “garbage in, garbage out,” where initial reliance on poorly vetted AI-generated materials—such as unverified training data, hastily generated specifications, or synthetic content—creates a foundation of misinformation that propagates and amplifies through each downstream process.

Example sentences:

The entire documentation suite was compromised by shit left—someone fed the LLM hallucinations into our requirements doc, and now every team is building against false assumptions.

We’re experiencing severe shit left in our content pipeline; the AI-generated summaries from sprint planning are so inaccurate that QA doesn’t even know what they’re testing anymore.

Related terms: garbage in, garbage out; technical debt; AI slop; shift left (antonym in intent)

See also: hallucination (AI), synthetic data poisoning, cascade failure