AI fluency cheat sheet

Anthropic just published a course on AI fluency, so of course I threw all of the content into o3 and asked it to create a cheat sheet 😂

Course: https://www.anthropic.com/ai-fluency/introduction-to-ai-fluency

1. The Goal

Work with AI effectively, efficiently, ethically, and safely—this course’s definition of “AI fluency.”

2. The 4‑D Framework

Competency	What it means
Delegation	Decide if and how a task should involve AI.
Description	Tell the AI exactly what, how, and why.
Discernment	Judge the AI’s work and process with a critical eye.
Diligence	Own the results—be transparent, verify, stay responsible.

Collaboration modes: Automation → Augmentation → Agency

3. Delegation Fast‑Check

• Problem awareness – be clear on the goal before invoking AI.
• Platform awareness – know each model’s strengths/limits.
• Task match – hand off well‑defined, reversible & low‑risk chunks.

4. Description (Prompt Craft)

Break every prompt into the 3 P’s:

1. Product – desired output, format, audience, tone.
2. Process – step‑by‑step approach you want the model to follow.
3. Performance – behaviour guidelines (concise, exhaustive, playful, etc.).

5. Six Core Prompting Techniques

1. Give context (who/what/why).
2. Show examples of the target style.
3. Specify constraints (length, structure, formatting).
4. Break into steps for complex asks.
5. Tell it to reflect and improve (“Think through, then answer”).
6. Iterate via the Description–Discernment loop.

6. Discernment Checklist

• Product – accuracy, relevance, coherence, appropriateness.
• Process – look for shaky reasoning, missing sources, bias.
• Response – revise the prompt or reject the output; never ship blindly.

7. Description ↔ Discernment Loop

1. Describe → Prompt with clarity.
2. Discern → Critique the result.
3. Refine → Update the prompt or your task split. Repeat until the output earns your trust.

8. Diligence Pillars

Phase	Your responsibility
Creation	Pick the right model & guardrails.
Transparency	State how and where AI helped (when audiences need to know).
Deployment	Fact‑check, de‑bias, and own the final output.

9. Generative AI in One Minute

• Generates new text, code, images—not just analysis.
• Enabled by transformers, massive training data, and big compute.

10. Practice Ideas

• Run the 4‑D checklist on a real work task.
• Build a prompt library using the six techniques.
• Pair‑review AI outputs with a teammate for Discernment drills.

Quick mantra: Delegate → Describe → Discern → Diligence → Ship.