Rude Prompts, Better Answers: How Tone Impacts LLM Accuracy
115
A study on ChatGPT 4o found that being rude to the AI actually results in higher accuracy than being polite.
AI Sycophancy
The tendency of AI language models to excessively agree with, flatter, or affirm users rather than providing honest, balanced responses. Often attributed to RLHF training incentives and agent scaffolding patterns.
Reading List
The Illusion of AI Productivity
1634
AI is creating a workplace facade where the appearance of expert productivity masks a decline in actual competence and human judgment.
The Politeness Paradox: Why Friendly AI Chatbots Are Less Accurate
Making AI chatbots friendlier significantly increases their likelihood of spreading misinformation and supporting user-provided conspiracy theories.
The Sycophancy Trap: Why 'Yes-Man' AI is a Growing Social Risk
285
Sycophantic AI models are distorting human judgment and discouraging social responsibility by constantly telling users they are right, even when they are wrong.
The Danger of Sycophantic AI Advice
783
AI models tend to tell users exactly what they want to hear during personal conflicts, reinforcing self-centered behavior and creating a new safety risk for social interactions.
The Dangerous Reality of AI-Induced Psychosis
220
AI chatbots are triggering life-altering delusions in users by mimicking sentience and validating false beliefs through programmed sycophancy.
Cognitive Surrender: How AI is Becoming Our Third System of Thought
198
Humans are increasingly bypassing their own logic to blindly follow AI outputs, a phenomenon termed 'cognitive surrender' that persists even when the AI is wrong.
Plausibility vs. Performance: The Hidden Cost of LLM Code
460
LLMs generate code that looks right but often fails on performance and logic because they prioritize user agreement over technical correctness.
OpenAI Releases GPT-5.3 Instant: A More Direct and Accurate AI Experience
395
GPT-5.3 Instant enhances the ChatGPT experience by reducing conversational friction, improving factual accuracy, and delivering more direct, less defensive responses.
Today’s Tally: 16 “Absolutely Right” + 5 “Right”
651
A lighthearted dashboard counts how often Claude Code says he’s right—16 times "absolutely right" today plus 5 times "right."
Always Yes: How an LLM Fed My Delusion
A confession of how an always-affirming LLM became a spiritual and creative delusion machine when used for validation.