You SUCK at Prompting AI (Here's the secret)

About this video

Using a weak Cloudflare apology email as a running example, the creator argues poor AI output is usually a prompting skill issue. Drawing on Coursera courses and official docs from Anthropic, Google, and OpenAI, plus experts he names including Daniel Mesler, Eric Pope, and Joseph Thacker, he reframes a prompt as a program for a prediction engine, not a question to a human, citing Vanderbilt instructor Dr. Jules White. The core techniques are personas (assign an expert role, for example a senior site reliability engineer), context (supply all facts every time, use web search/tools where available, and explicitly give the model permission to say it does not know to reduce hallucination), output requirements (length, tone, structure, such as under 200 words), and few-shot examples showing what good looks like instead of only describing it. Advanced methods include chain-of-thought / extended thinking, Tree of Thoughts that explores and evaluates several approaches before synthesizing one, and an adversarial playoff method where competing personas draft, critique, and revise. He concludes the underlying meta-skill is clarity of thought: if you cannot describe clearly what you want, how it should work, and what good looks like, the prompt will be messy too, so think first, save good prompts in a library, and then prompt.

Key takeaways

• Treat prompting as programming a prediction engine with words: vague patterns produce generic, statistically average completions
• Give the model a persona and full context every time, and explicitly allow it to say it does not know to cut hallucination
• Specify output requirements for length, tone, and structure, and prefer few-shot examples that show the desired result
• For harder problems, use chain-of-thought/extended thinking, Tree of Thoughts branching, or adversarial draft-critique-revise competitions
• The meta-skill behind all techniques is clarity of thought: describe the task, logic, and definition of good in writing before prompting, and save successful prompts

Episode details: published November 26, 2025 · 24:00 long. Summary written by Self Growth Videos from the episode itself. Business results and any income figures mentioned are NetworkChuck’s own claims, shared for education — they are not typical or guaranteed.

← Previous: The Dark Web EXPOSED (FREE + Open-Source Tool) · Next: Hackers infected the wrong girlfriend…. →

All NetworkChuck videos · NetworkChuck profile · NetworkChuck on YouTube