Everyone is Getting AI Fluency Wrong—Steal My 10 Level Framework That Exposes the Real AI Skill Gap
Most AI skill assessments are anchored to specific tools or job verticals — how well do you use ChatGPT, how effectively can you prompt as a developer. Nate B Jones argues this misses the point entirely. What matters is not which AI product you use but how deeply you understand the principles underneath all of them. His framework is model-agnostic, platform-independent, and designed to be evergreen as the technology evolves.
The core insight: roughly 80% of AI users today fall between Level 1 and Level 5 on his scale, mostly doing transactional tasks with no real mental model of how LLMs work. The gap between that majority and the people operating at Levels 5–9 — who think in systems, optimize for prompt yield, and actively discover undocumented capabilities — is where the real AI skill gap lives. This framework maps where you are, what the next level looks like, and how to think about your own development as the baseline keeps shifting.
The AI Fluency Scale
The scale runs from 1 to 9. There is no Level 10 — intentionally. The technology is continuously evolving, so the top of the scale remains open. The point is not to reach some terminal state but to understand where you are relative to where you want to be.
1. Levels 1–3: The Basic Beginner
This is the default starting point for the vast majority of AI users. If you use ChatGPT, Copilot, or Claude primarily to rewrite emails, summarize text, or make adjustments to documents, you are in this range.
Jones emphasizes this is not a judgment. The framework exists to help you understand your current position so you can figure out where you want to go. Not everyone needs to be a 10 — that is not the point. The point is understanding your level, defining your goals, and making sure you are equipped to get there.
At this stage, interaction with AI is intuitive but shallow. Users get results without understanding how the tool processes their request or why it produces what it does.
2. Levels 3–5: Building Mental Models
The defining shift at this stage is the development of a mental model for how LLMs actually work. Jones considers this the most important transition on the entire scale, and the one that most people talk about least. Most AI education jumps straight to specific skill sets. He argues you need an overarching perspective on fluency and competency first.
Understanding LLM mechanics. Users at this level recognize that:
- LLMs do not “know” things in a human sense
- They are not traditionally programmed
- Next-token prediction is the core mechanism
- There is a meaningful difference between when an AI is reasoning and when it is not
Context retrieval. As AI has gotten more powerful, understanding context retrieval has become essential at this level — more so than it used to be. These models can now take book-sized context windows, and users need a mental model for how that works. Importantly, this does not mean you can build a RAG system or a memory system. If those terms are over your head, you can still be firmly at Level 3–5 as long as you have the conceptual mental model down.
Outcome-driven prompting. This conceptual understanding naturally leads to a shift in how users approach prompts. Instead of asking “what should I tell the AI?”, they start asking “what is the output I need?” The mental models inform their understanding of how outputs are created, and they begin working backward from the desired result. Jones calls this “intuitive prompt engineering” — you are not reading from a book or copying prompts. Maybe you use templates sometimes, but you know how to massage and tailor them. You can write prompts yourself because you understand how to get to the outcome you want.
3. Levels 5–7: Systematization and Professional Use
If you get above Level 5, you are treating AI as a serious professional tool. The defining characteristic at this stage is systems thinking. Patterns start to come through that simply do not appear at lower fluency levels.
Auditable patterns. Users transition from casual, intuitive prompting (“I usually do this”) to reliable, predictable sequences (“this is the sequence I follow, I get a predictable result, and I can systematize it in a way that others can do it too”). The difference is moving from intuition to understanding the system well enough to predict and move with it.
Prompt yield. This is the quality of output per unit of prompting. An inefficient prompter might take 10 iterations to get one usable output. An efficient one does one or two prompts and gets 98% of the way there. Jones emphasizes that valuing tokens and valuing time is a hallmark of this level — doing the prompt correctly to get the right answer, rather than iterating your way there.
The shift from Level 3–5 to Level 5–7 on prompt yield looks like this: moving from “I think this is the right prompt to get this output” (casual, intuitive) to “this is the yield I get on this prompt, I think it can be modified in these three ways, and I’m going to get a much more efficient output” (systematic). Then they make the change, measure the result, and confirm. These people think in feedback loops.
Tool stacks and peer leadership. Most people at this stage will have:
- A personal prompt library
- 5–7 AI tools they work with regularly
- Preferences for specific work tasks associated with those tools
- Recognition from teammates as a peer collaborator who can help the team put systems in place
The generalist conviction. Jones holds a strong conviction that AI is a generalist skill set, and the industry is probably teaching it wrong by diving too deep into verticals without a generalist conceptual foundation. It is great to understand how to build with LangSmith as a developer, but that is not the only kind of AI learning you need. The general approach to skill sets and fluency — this common understanding — is the missing piece.
4. Levels 7–9: Teaching, Innovation, and Discovery
At this point, you have mastered systems thinking and possess a deep conceptual understanding of LLMs. You are a teacher and a trailblazer.
Teaching as a learning mechanism. Jones speaks from personal experience: teaching has been instrumental in driving clarity and revealing gaps in his own understanding that he has to relentlessly close. Most teachers, regardless of subject, will confirm this. At this level, you should think about who you can teach and how your teaching drives your own growth.
Documentation and scaling influence. The more you document what you are learning, thinking, and discovering, the more you can scale your influence and teach others. “It’s not about growing influence,” Jones clarifies. “It’s about being able to communicate really clearly things that are net new in the space that you can then understand how to teach others in a way that is accessible for their level.”
This might look like:
- Setting up the AI training curriculum at work
- Leading a group of developers through their first AI build
- Creating public content — YouTube, Substack, shared project templates, evaluative tools
Discovering capabilities. People who understand LLMs deeply know that their capabilities are not fully documented at release. It is more accurate to say these systems are “grown” than to say they are programmed. OpenAI does not ship an AI and know everything about it. The community collectively discovers what grew. At Levels 7–9, part of the job is to innovate — to understand where to push farther on LLM capability, why it matters, and then turn around and teach that back to grow the practice.
Pulling the impossible into the possible. The innovation piece at this level means helping to pull forward things that were previously deemed very difficult to do with AI, because you are actively discovering and putting undocumented capabilities to new uses.
The Shifting Baseline
Jones deliberately does not assign a Level 10. Instead, he frames the entire scale against a moving baseline.
The competitive reality is shifting. As we move from 2025 into 2026 and beyond:
- The general population will naturally grow into the Level 1–3 range
- A much larger demographic will push into Levels 3–5
- Many more people will be climbing the skill ladder above that
- The skills required at each stage will continuously evolve
This is not meant to create panic. Your goal may not be to become a teacher or instructor — maybe you want to be a systems thinker, or maybe you are perfectly happy just understanding how LLMs work. But regardless, the skills required at each level are evolving.
Jones’s advice: “Think of it as a moving train, and it’s never going to go slower than it’s going right now.” Board the train and progress at a pace aligned with your career path and goals. Develop a mental model not just of AI, but of your own career path — have a sense of what level of fluency would be useful in your job family.
Applying the Framework to New Technologies
The extra credit, and the reason Jones calls this an evergreen framework: as new technologies launch, you can map them directly onto this scale to understand how different fluency levels will interact with them.
Example: The launch of AI agents (Jones references OpenAI’s agent framework launch on October 6, 2025):
- Levels 1–3: Will find the concept of autonomous agents confusing or overly difficult — “this agent thing looks really hard”
- Levels 3–5: Will seek intuitive ways to use a single agent to accomplish a real-world task, expressing their mental model of LLMs to get real work done
- Levels 5–7: Will focus on systematizing the technology — building not just one agent but multiple interacting agents, and sustaining them efficiently within an organization
- Levels 7–9: Will innovate, discovering entirely new and undocumented ways to utilize agent frameworks, then teaching those methods to the broader community
This mapping works for any major AI technology launch going forward. The framework is not an October 2025 artifact — it is a living tool for making sense of your own skill level relative to where the technology is, one you can return to again and again.
“I wanted to put together a sense of the jungle gym of AI and a sense of where people can go. And then hear where all of you want to get to.”
Meta
Added: 2026-04-13