Original Insight
“You have a problem, like weight loss. You can tinker on that level as much as you can. But then you can ask, what causes the problem? And is there a deeper level than that? So you can go 3 to 4 levels down. And it’s a completely different context that approaches it… Then I combined that with recursive prompting within that level, and on top of that, first-principles priming. When you run an article through this process, just publish it, because it will have such good quality.” — Kasimir
Expanded Synthesis
Most people treat AI prompting like a search engine upgrade — ask the surface question, get a surface answer, move on. What Kasimir described in the February 5th session is something categorically different: a layered, multi-level prompting architecture that forces the AI to traverse the full depth of a problem rather than skim its obvious surface.
The structure he outlined works like this. You present a topic — let’s say weight management, or executive decision-making, or client retention. Rather than immediately prompting at that level, you first ask the AI: what are the underlying causes or mechanisms behind this topic? Then you go again: is there a layer deeper than that? After two or three iterations of this vertical descent, you have surfaced the structural roots of the problem rather than its symptoms. Only then do you introduce recursive prompting — asking the AI to refine within that level — combined with first-principles priming, which means asking: what are the founding principles related to this specific layer we’ve surfaced?
This is the difference between coaching a client on their surface behavior (the tactic) and understanding the cognitive and emotional architecture underneath (the system). A coach who only ever works at the surface level will exhaust their clients with a constant stream of new tactics that don’t stick, because the underlying structures haven’t shifted. The same dynamic applies to AI-assisted work: prompts that operate at the surface produce generic outputs that could have been written by anyone. Prompts that descend three or four levels produce outputs that are structurally differentiated — they address things the median person isn’t thinking about.
For knowledge entrepreneurs and coaches specifically, this matters enormously. The content that builds genuine authority is not the content that restates what everyone already knows. It’s the content that illuminates mechanisms people didn’t realize were operating in their own lives. A three-level descent on “why executives resist delegating” will surface entirely different content than a surface-level prompt on “delegation tips for leaders.” The first addresses identity, psychological safety, and the hidden costs of perceived incompetence in peers. The second produces a listicle.
There is also a structural efficiency here that Kasimir noted: when you arrive at an article through this process, the quality is high enough that it can be published with minimal editing. This is not because the AI did everything — it’s because the depth and structure of the input produced depth and structure in the output. The human judgment about which level to descend to, when to stop, and how to frame the first-principles priming is the non-replicable contribution.
For PowerUp clients — coaches and high-performers who are building thought leadership — this is a direct competitive advantage. The market is flooded with AI-assisted content that was produced by people asking first-level questions. The scarcity is in depth. Depth is produced by human curiosity operating through a rigorous process, not by better tools.
The psychological mechanism at work here is also worth naming: most people stop at level one because they feel like they’ve understood the problem once they can name it. Naming is not understanding. Naming a challenge gives the illusion of comprehension while obscuring the mechanism. This is the curse of labeling in coaching: the moment a client says “I have an imposter syndrome problem,” they feel they understand it — and simultaneously lose access to the more interesting questions underneath. Multi-level prompting is, in part, a discipline for breaking through premature naming.
Practical Application for PowerUp Clients
The Three-Level Descent Exercise
Use this whenever you’re creating content, coaching a client, or solving a complex problem with AI assistance.
-
Level One — Name the Surface Problem. What is the obvious issue or topic? Write it in one sentence.
-
Level Two — What Causes the Problem? Ask yourself (or prompt the AI): what are the underlying mechanisms, beliefs, or structural conditions that produce this surface issue? Push for at least three distinct answers.
-
Level Three — What Produces Those Causes? Go one level deeper. What systemic or psychological architecture generates the causes you just identified? What would need to be true at a root level for those causes to persist?
-
First Principles Priming. At the level you’ve arrived at, ask: what are the foundational principles, the first principles, of this domain or problem type? Now prompt the AI from this level.
For Content Creation: Run a topic you’re writing about through this descent before you write a single word. Use the level-three framing as the controlling idea of your piece.
Coaching Questions:
- “If we peel back one more layer — what’s driving that pattern?”
- “What has to stay true in your life for this problem to persist?”
- “What would be different in your identity if this were no longer a constraint?”
Additional Resources
- Thinking in Systems by Donella Meadows — the foundational text on structural vs. symptomatic thinking
- The Fifth Discipline by Peter Senge — systems archetypes as a framework for multi-level problem analysis
- Insight - Codify Your Judgment Into Skills, Not Just Prompts — the companion move: once you develop a multi-level process, turn it into a repeatable skill
Evolution Across Sessions
This insight from February 5th forms an important precursor to the eigenthinking framework Lou introduced in the February 19th session. Both are fundamentally about using structured processes to push AI — and human thinking — past the median answer into the latent space where genuinely differentiated insights live. The multi-level descent is the entry method; eigenthinking is the navigation system within that space.
Next Actions
- For Lou: Consider developing a “depth audit” coaching exercise based on the three-level descent — clients apply it to their current biggest problem before a session, and bring the level-three framing to the call. This creates richer material and builds the habit of structural thinking.
- For clients: Assign a three-level descent on one key content topic per week. Publish the resulting article as a test of depth against their current output.