Jonathon's AI Wiki

Winston MIT Presentation Prompts (God of Prompt)

4 min read

Source: Godofprompt Patrick Winston Mit Presentation Prompts — thread by @godofprompt on X, 2026-04-01, extracted via browser automation 2026-04-12.

A six-prompt Claude workflow by @godofprompt that operationalizes Patrick Winston’s MIT “How to Speak” framework end-to-end — opening, slide design, memorability, talk structure, teaching with props, and closing. Every prompt uses the same <role>, <task>, <steps>, <rules>, <output> XML-tag skeleton, making it a clean reference implementation of Anthropic’s structured-prompt conventions applied to a single domain.

Key Takeaways

  • Consistent XML skeleton — every prompt in the thread uses <role> / <task> / <steps> / <rules> / <output>. The repetition itself is the lesson: a good prompt template is reusable across tasks.
  • Role goes past “you are a…” — each <role> names a specific stance (presentation coach, slide crime investigator, personal brand architect, persuasion architect, teaching design specialist, closing specialist) and binds it to a named framework (Winston’s empowerment promise, 10 slide crimes, Star framework, job talk structure, prop/storytelling, contributions close).
  • Steps start with “ask me first” — every prompt’s step 1 is “ask for my topic / audience / goal before starting.” This is the interview pattern from Prompt Engineering Essentials applied consistently.
  • Rules are negative constraints with specific alternatives — “never open with a joke,” “never end with thank you,” “font minimum 40pt.” Each ban names what to do instead. This is the explicit-criteria pattern, not vague guidance.
  • <output> defines deliverable shape — a short arrow-delimited sequence like Empowerment Promise → First 60 Seconds → What to Cut → Opening Script. Predictable structure without schema enforcement.
  • Winston’s Star mnemonic (prompt 3) — Symbol, Slogan, Surprise, Salient idea, Story. Worth capturing as a concept outside the prompt.

The Six Prompts

Full prompt text lives in the source file: Godofprompt Patrick Winston Mit Presentation Prompts. Summary:

  1. Start Any Presentation Right — empowerment promise, first 60 seconds, what to cut.
  2. Eliminate Your Slide Crimes — audit against Winston’s 10 crimes (too many slides, font under 40pt, “thank you” final slide, etc.), fix each, redesign closing as contributions slide.
  3. Make Your Ideas Unforgettable — apply Winston’s Star: Symbol → Slogan → Surprise → Salient idea → Story.
  4. Structure Any Talk That Persuades — vision, proof of work, contributions close, all within first 5 minutes.
  5. Use Props and Stories to Teach Anything — physical prop, story arc with tension, verbal script guiding attention.
  6. End Any Presentation Powerfully — contributions slide stays during Q&A, audience salute, no “thank you” or “questions?” slide.

The Prompt Template

Every prompt in the thread fits this shape — portable to any domain:

<role>Act as a [named specialist] applying [named framework] — [one-sentence philosophy].</role>

<task>[Single concrete deliverable the user needs.]</task>

<steps>
1. Ask for [key context] before starting
2-5. [Named sub-steps in execution order]
</steps>

<rules>
- Never [banned behavior] — [specific reason]
- [Positive constraint with measurable criterion]
- [Constraint naming alternative behavior]
</rules>

<output>[Arrow-delimited deliverable sequence]</output>

Try It

  • Copy one prompt verbatim and run it against a real upcoming presentation. The “ask me first” step means Claude will interview you before producing output — let it.
  • Adapt the template to a non-presentation task you repeat often (code review, RFC write-up, interview prep). Keep the five-tag skeleton; swap the Winston framework for your own named methodology.
  • Build a /winston-presentation slash command in Claude Code that chains all six prompts in sequence with the interview questions front-loaded. See Claude Code Subagents for the execution model.
  • Extract Winston’s Star (Symbol / Slogan / Surprise / Salient idea / Story) as its own framework reference. It’s reusable beyond presentations — product positioning, blog headlines, ad copy.

Open Questions

  • Does this six-prompt chain work better as six separate runs or as one chained super-prompt? The thread doesn’t test the alternative.
  • How much of the output quality comes from Winston’s framework vs the prompt template? Would the same template with a different framework (e.g., Nancy Duarte’s “Resonate”) produce comparably structured results?
  • The <output> tag defines only the delivery shape, not a schema. Would adding JSON-schema tool_use calls improve consistency across runs, or over-constrain the creative output?
  • Thread claims Winston’s framework was taught “for 40 years” — Winston’s “How to Speak” lecture is well documented, but the specific 10-slide-crime list and Star framing attributions should be verified against the primary source before citing as authoritative. ^[inferred]

Graph View

Backlinks