---
id: "prereq-transformer-architecture"
type: "prereq"
source_title: "Don't Let AI Slop Muck Up Your Company's Processes"
source_url: "https://hbr.org/2026/06/dont-let-ai-slop-muck-up-your-companys-processes"
source_timestamps: ["¶17", "¶18"]
tags: ["machine-learning", "llm-mechanics"]
related: ["concept-knowledge-entropy"]
reason: "Necessary to understand why knowledge entropy and hallucinations are fundamental architectural features of current AI, not just temporary bugs."
sources: ["execution"]
sourceVaultSlug: "hbr-seg-execution"
originDay: 8
articleStem: "hbr-sig-54-ai-slop-processes"
sourceUrl: "https://hbr.org/2026/06/dont-let-ai-slop-muck-up-your-companys-processes"
sourceTitle: "Don’t Let AI Slop Muck Up Your Company’s Processes"
---
# Transformer Architecture Basics

**What to know.** LLMs are built on transformer algorithms — context-agnostic statistical models that produce probabilistic, next-word-prediction output. They have no intrinsic conception of fact or truth.

**Why it matters.** Necessary to understand why [[concept-knowledge-entropy]] and hallucinations are fundamental architectural features of current AI, not temporary bugs. This is why [[quote-llm-entropy]] states entropy can be *managed but not eradicated* short of a step-change in architecture, and why [[concept-generative-inbreeding|model collapse]] is a structural risk rather than an implementation defect.