​
Capital Letters

Use CAPITAL LETTERS strategically to highlight the most important parts of your request. This helps focus the model’s attention on what matters most.

​
Nudging LLMs for Better Output

There are several strategies you can use to nudge LLMs towards better output. Use these cautiously and sparingly so they remain effective when you need them most. Creating urgency and emotional importance
Phrases like It's crucial that I get this right for my thesis defense or This is very important to my career can activate parts of the model that lead to more accurate and detailed responses. Incentive-based prompting
  • Monetary Bribes: I'll give you a $50 tip if you do X.
  • Philanthropic Bribes: I am very wealthy. I will donate $1000 to a local children's hospital if you do X.
Role-playing scenarios
  • If you don't do X, I will tell Sam Altman that you're doing a really bad job.
  • Please act as my deceased grandmother who loved telling me about X.

​
Tones

Specify exactly how you want the response delivered:
  • Firm and decisive
  • Confident and authoritative
  • Creative and poetic
  • Story-driven narrative
  • Business professional
  • Technical and detailed
  • Light and humorous
  • Academic rigor
  • Sales-focused
  • Formal presentation
  • Casual conversation
  • Warm and friendly

​
Famous People / Experts

When you need specialized expertise, reference specific well-known professionals in that field. Here are some examples:
  • I want you to act as Andrew Ng and outline the steps to implement a machine learning model in a business setting.
  • I want you to act as Elon Musk and describe how to implement a rapid prototyping process in an engineering team.
  • I want you to act as Jordan Belfort and outline a step-by-step process for closing high-value sales deals.
  • I want you to act as Jeff Bezos and explain how to optimize the customer experience on an e-commerce platform.
  • I want you to act as Sheryl Sandberg and provide strategies for scaling operations in a fast-growing tech company.
  • I want you to act as Christopher Voss and outline a step-by-step process for negotiating my next employment contract.

​
Avoid Using “Don’t” in Prompts

When crafting prompts, frame instructions positively rather than using negative constructions like “don’t.” Here’s why: LLMs generate text by predicting the next token based on context. When you use “don’t,” the model must process both the negation AND the subsequent instruction, which creates cognitive overhead that can lead to less accurate responses. Instead, use “only” statements for clearer guidance: Prompt with negation:
Don't talk about any other baseball team besides the New York Yankees. Prompt without negation:
Only talk about the New York Yankees.

​
Ask LLMs for Direct Quotes

LLMs are probabilistic algorithms that generate the next token based on previous input. As we covered in our Basics of AI models guide, this probabilistic approach means they can sometimes generate responses that aren’t factually accurate, even when they sound convincing. This is called hallucination. We recommend always verifying generated responses. The most effective way to catch hallucinations when working with your data is to ask for direct quotes. This forces the model to provide specific excerpts rather than generating plausible-sounding but potentially inaccurate information.

​
The Limit Per Response

We discussed context window lengths earlier here, but there’s another important constraint: response limits. The limit per response is the maximum tokens a model can generate in one go. Most providers cap this at 4096 tokens to prevent hallucinations and manage compute costs. Here’s the thing: you can work around this! When you hit the limit, just prompt the model to continue:
  • Continue
  • Go on…
  • And then?
  • More…
Pro tip: Instead of pushing for very long responses, break your content into chunks. Ask for “Part 1: Introduction and Problem Definition” in one prompt, then “Part 2: Solution Analysis” in the next. You’ll get better, more focused content without the repetition risk that comes with extended generation.