/

/

Back to Learning Paths

IntermediatePrompt Engineering

Mental Models: How to Think with LLMs

This path helps you shift from using prompts to thinking like a prompt engineer. Instead of teaching techniques, it introduces powerful mental models — ways of seeing and reasoning about large language models. Each guide is a mindset essay with visual diagrams and reflection prompts that sharpen your intuition, inspire creativity, and build the resilience needed to invent new solutions. By the end, you’ll think like a builder, a critic, and a collaborator with AI.

20

Guides

160

Duration

6

Skills

Learning Path Content

Step 1Intermediate8 min read

LLM as a Stochastic Parrot

LLMs don’t think: they predict. Treat them as stochastic parrots: shape probabilities, embrace variability, and turn randomness into leverage.

8 min read

Step 2Intermediate8 min read

LLM as a Simulator of Worlds

See LLMs as simulators of worlds: define roles, rules, and settings, then watch coherent scenes unfold.

8 min read

Step 3Intermediate8 min read

LLM as a Crowdsourced Brain

See LLMs as a chorus of voices: consensus, not truth. Learn to convene, separate, aggregate, and filter for stronger answers.

8 min read

Step 4Intermediate8 min read

LLM as a Probabilistic Compiler

See LLMs as probabilistic compilers: prompts are source code, outputs are text-executions. Clear specs = fewer bugs.

8 min read

Step 5Intermediate8 min read

LLM as a Glass Box, Not a Black Box

LLMs aren’t black boxes: you can ask for receipts. Answers, evidence, and process make outputs auditable and trustworthy.

8 min read

Step 6Intermediate8 min read

LLM as an Adversary-in-Waiting

Treat LLMs as eager but unsafe actors—assume breach, contain agency, and let outputs earn trust before execution.

8 min read

Step 7Intermediate8 min read

LLM as a Puzzle-Solver with Bias

LLMs rush to “solve” puzzles, but often the wrong ones. Frame objectives and constraints so its bias works for you, not against you.

8 min read

Step 8Intermediate8 min read

LLM as a Compression Engine of Human Knowledge

See LLMs as compression engines: fluent but lossy. Learn to rehydrate answers for the right level of fidelity.

8 min read

Step 9Intermediate8 min read

LLM as a Probability Map, Not a Straight Line

See LLMs as probability maps, not straight lines: explore multiple routes, compare, and choose the best.

8 min read

Step 10Intermediate8 min read

LLM as an Alien Colleague

Work with LLMs as if they were alien colleagues: brilliant but with different priors. Translate intent, set boundaries and build shared footing.

8 min read

Step 11Intermediate8 min read

LLM as a Tool Orchestrator

LLMs act as conductors, coordinating external tools to create precise, reliable outcomes. Design the orchestra, not just the soloist.

8 min read

Step 12Intermediate8 min read

LLM as a Probability Trickster

LLMs speak with confidence, even when wrong. Learn to separate fluency from truth and build the reflex: trust, but verify.

8 min read

Step 13Intermediate8 min read

LLM as a Cognitive Multiplier

LLMs don’t replace thinking: they amplify it. Sharper framing leads to sharper answers and better decisions.

8 min read

Step 14Intermediate8 min read

LLM as a Simulator of Context

LLMs don’t “know,” they simulate worlds. Shape the scene, world, persona, process and the model projects the next plausible moves.

8 min read

Step 15Intermediate8 min read

Prompts as Interfaces, Not Commands

Prompts aren’t magic spells—they’re interfaces. Treat them as protocols with inputs, outputs, and rules for consistency and reliability.

8 min read

Step 16Intermediate8 min read

LLM as a Jury, Not a Judge

Think of the LLM as a jury: it offers multiple voices, not a single ruling. Your job is to weigh, compare, and decide.

8 min read

Step 17Intermediate8 min read

LLM as a Trickster

Work with the LLM as a confident trickster: creative but fallible. Generate, then verify.

8 min read

Step 18Intermediate8 min read

Failure as Fuel

Turn every bad output into training data. Failure isn’t wasted: it’s feedback that fuels sharper prompts and stronger results.

8 min read

Step 19Intermediate8 min read

LLM as a Multiverse Explorer

Explore LLMs as multiverse explorers: each run a different path. Learn to fan out, weigh, and converge on the best outcome.

8 min read

Step 20Intermediate8 min read

LLM as a Mirror of Humanity

An LLM is a mirror reflecting human patterns; polish data, prompts, and perspective to shape better reflections.

8 min read

Start Learning Path

Path Information

Prerequisites:

- Basic familiarity with LLMs and prompt engineering (Beginner Path recommended) - Comfort reading analogy-driven essays - No coding or technical setup required

Duration:

160

Level:

Intermediate

Skills You'll Learn

Systems thinking for LLM behaviorAdversarial & out-of-the-box thinkingPrompt design intuition beyond recipesRecognizing model limitations and biasesCreativity and persistence in problem framingTranslating mental models into practical prompting strategies

Learning Outcomes

Reframe LLMs as probability machines, not oracles, and design prompts accordingly
Adopt adversarial and hacker-like thinking to stress-test their own prompts
Use metaphors (mirror, actor, negotiator, amplifier) to inspire new solutions
Recognize failures as data and turn them into iterative improvements
Balance creativity with control by thinking of prompts as interfaces, not commands
Develop the persistence and mindset to invent their own advanced prompting patterns
See LLMs as collaborators, not tools — amplifying their own framing and imagination