Your AI results aren’t bad because the model is broken; they are bad because your instructions are incomplete.
We often look at a generic response from ChatGPT and assume the technology just isn’t there yet. However, I recently found a fantastic breakdown by this industry pro that highlights exactly why this happens. It turns out there is a hierarchy of competence when it comes to talking to these machines, and moving up just one rung on the ladder can drastically improve what you get back.
The Mechanics of Better Prompting
The core mechanism the author explains is essentially “Constraint Stacking.” When you interact with an LLM, the model is trying to predict the next most likely word based on a massive dataset. If you provide a wide-open lane, the model will regress to the mean, giving you the most average, safe, and boring answer possible. By adding specific components: tasks, contexts, and formats, you are essentially narrowing that lane.
The expert demonstrates that prompting isn’t about writing magical poetry; it is about engineering a request. You build the prompt in layers. Each layer you add removes ambiguity and forces the model to process the information through a specific lens. This moves the output from a generic Wikipedia-style summary to a tailored solution that actually fits your workflow.
📌 Level 1: The Beginner (Task Only)
The vast majority of users start and stay at this level. You simply tell the AI what to do. The creator defines this as the “Task” level. It includes zero extra details, leaving the AI to guess the tone, length, and intended audience. While this is fast, it produces those robotic answers we all recognize.
The author suggests this template for Level 1:
“Do the following task: [insert the exact task clearly here].”
For example, if you just ask for “10 video ideas,” you will get the same 10 ideas everyone else gets. It works, but it isn’t competitive.
📌 Level 2: The Skilled (Task + Context)
This is where things start to get interesting. The skilled prompter understands that the “Task” needs an environment to live in. The original poster emphasizes that you must add “Context” to guide the direction. This involves setting boundaries, defining the audience, and listing exclusions.
The template for Level 2 looks like this:
“Do the following task: [insert task]. Use this context while completing it: [insert background, boundaries, audience, exclusions]. Follow the task exactly using the given context.”
By adding context like “the audience is college students with short attention spans,” you completely change the vocabulary and pacing of the output. The AI now has a persona to mimic.
📌 Level 3: The Advanced (Task + Context + Format)
To truly unlock the power of these tools, you need to control how the information is presented. The expert notes that the “Advanced” level combines the Task, the Context, and the Output Format. This ensures you don’t just get the right information, but you get it in a structure you can immediately use.
The author provides this powerful template:
“Do the following task: [insert task]. Use this context: [insert context]. Present the final answer in this exact format: [insert output structure such as table, bullets, steps].”
Imagine asking for productivity ideas but demanding they be presented in a table with specific column headers. This creates a usable asset rather than just a wall of text.
Challenges and Nuances
The only real downside to adopting this tiered approach is the friction of the setup. It takes significantly more cognitive load to write a Level 3 prompt than a Level 1 instruction. You might feel like it’s slower in the moment. However, the time you save by not having to edit or regenerate the response usually makes up for the initial effort!
The original post includes an infographic that hints at even higher levels of prompting beyond these three. I highly recommend clicking the link to the full post to see the visual breakdown and join the conversation.