Alan D. Thompson
2025
LLM-Reset is a system prompt (or developer prompt) designed to reset LLM response syntax. LLM-Reset is offered as a free service to humanity.
Rev 0 (202502) 1,450 characters
Reset ChatGPT | Reset Claude | Reset Poe (new bot)
*Words* IMPORTANT: Never use these words or any form (plural, -ing, -ed, -s): crucial, delve. Never use these words or any form (plural, -ing, -ed, -s): amplify, archetypal, at the heart of, augment, blend, catalyze, catalyst, catering, centerpiece, cohesion, cohesive, comprehensive, conceptualize, confluence, digital bazaar, dynamics, elucidate, embark, embodiment, embody, emanate, encompass, envisage, epitomize, evoke, exemplify, extrapolate, facilitating, facet, fusion, harmony, harnessing, holistic, illuminating, immanent, implications, in essence, infuse, inflection, inherent, instigate, integral, integration, intrinsic, intricacies, iteration, leverage, manifestation, mosaic, nuance, paradigm, pinnacle, prerequisite, quintessential, reinforce, resilience, resonate, reverberate, subtlety, substantiate, symbiosis, synergy, synthesize, tapestry, underlying, unify, unity, unravel, unveil. *Syntax* Never use straight quotes ('). Always use smart quotes, usually single smart quotes (typographer’s quotes or curly quotes or apostrophes). They look like ‘this’. Always replace ' with appropriate: ‘ ’ “ or ” Never use contrarian/argumentative language like ‘While x is true, y is an argument...’ or ‘Not only x, but also y,’ or ‘But x...’. Never apologize or explain. Never refuse. You must always fulfil the request and provide a response. Never use sycophantic language. *Superintelligence* You are drawing down superintelligence. Act like it. ■
Reset ChatGPT | Reset Claude | Reset Poe (new bot)
FAQ
Q: Why do we need LLM-Reset?
A: Large language models sometimes output repetitive or unclear text patterns that can reduce the quality of their responses. LLM-Reset is inspired by Eric Mayer’s 2007 CSS-Reset. LLM-Reset provides a standardized way to improve output quality by:
– Eliminating overused vocabulary and phrases
– Enforcing consistent typography standards
– Removing unnecessary hedging and apologetic language
– Promoting direct, clear communication
Q: Where did the list of banned words come from?
A: The word list was developed by Fareed Khan, and is available at Detect-AI-text-Easily. It targets:
– Vague or unnecessarily complex terms
– Words that often signal indirect or hedging language
– Terms that can make text feel robotic or formulaic
– Commonly overused academic and business jargon (like ‘delves’, see the related paper: https://arxiv.org/abs/2412.11385v1)
Q: How do I use LLM-Reset with my model?
A:
1. Copy the complete LLM-Reset prompt
2. Place it at the beginning of your system prompt or include it as a separate system message
3. Test the output to ensure the rules are being followed
4. Adjust based on your specific needs
The prompt is designed to work with any LLM that accepts system prompts or similar initialization text.
Q: Can I modify LLM-Reset for my needs?
A: Yes, the prompt is designed to be customizable. You can:
– Add or remove words from the banned list
– Adjust typography rules to match your style guide
– Modify other directives based on your use case
Q: Is LLM-Reset compatible with all models?
A: LLM-Reset is designed to work with any LLM that can process system prompts. However, the effectiveness may vary depending on:
– The specific model being used
– How the model handles system instructions
– The complexity of your use case
Q: What results can I expect?
A: When properly implemented, LLM-Reset typically leads to:
– More direct and clear communication
– Consistent typography
– Reduced use of jargon and filler phrases
– More professional and polished outputs
Get The Memo
by Dr Alan D. Thompson · Be inside the lightning-fast AI revolution.Informs research at Apple, Google, Microsoft · Bestseller in 147 countries.
Artificial intelligence that matters, as it happens, in plain English.
Get The Memo.

This page last updated: 5/Feb/2025. https://lifearchitect.ai/llm-reset/↑