Never Trust An LLM @mattpocockuk
Never Trust An LLM  @mattpocockuk
Uploaded March 2026 | Updated November 2026, 6 days ago
LLMs hallucinate constantly - they lie about facts, invent entities, and ignore context. Learn why this happens, how it works, and how to protect yourself from AI misinformation.

0:00 LLM's Are Liars
1:33 Factual Errors
3:22 Fabricated Entities
4:04 Contextual Inconsistency
5:24 LLM's Are Compressors
7:08 Why Do They Guess?
10:01 How To Fix It

LLM Taxonomy:
arxiv.org/abs/2508.01781

Why Language Models Hallucinate:
arxiv.org/abs/2509.04664

Join the Claude Code for Real Engineers course:
https://www.aihero.dev/cohorts/claude-code-for-real-engineers-2026-04

Sign up for the AI Hero newsletter:
https://aihero.dev/newsletter

Follow Matt on Twitter

twitter.com/mattpocockuk

Join the Discord:

https://aihero.dev/discord
Never Trust An LLMNever Run claude /initTell your agents to CARE about the codebaseKill your MEMORY.mdWill the art of coding disappear?New Skills! v1.1 brings /wayfinder, /research, /implement, /to-spec, /to-ticketsYes, LLMs can produce more than just textLibrary types can run, but they cant hideThe 7 phases of AI-driven developmentMost devs dont understand what agents areYour types cannot hide from the wizard #3console.dir, where have you been all my life
Matt Pocock |

Never Trust An LLM

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER