authority.md vs Writing Your Own Prompts: Which Actually Works?

The most common "competitor" to authority.md isn't another marketplace. It's the voice in your head saying "I could just write this myself."

You could. A technically competent prompter can ask Claude to "think like Warren Buffett when evaluating this investment" and get something usable. This post is an honest comparison of that DIY path versus paying $4.99 for an engineered framework.

The answer isn't universal. It depends on what you're doing, how often you'll use it, and how much your time is worth.

The honest premise

Our homepage literally says "Yes, you could prompt this yourself. We spent months engineering it properly." We aren't pretending it's impossible to DIY. We're selling the engineering.

This post shows you what that engineering actually looks like and when it's worth paying for.

The 5-minute prompt most people write

Here's what a competent prompter types into Claude:

Act as Warren Buffett — value investor, long-term thinker, focus on businesses with economic moats. Help me evaluate [X].

You'll get a reasonable Buffett-flavoured response. The AI adopts the voice, mentions moats, talks about margin of safety. Output quality: about 60% of what's possible.

When this is enough:

• One-off use
• Casual exploration
• You don't mind re-writing the prompt each time
• You're happy with "sounds like Buffett" output

The 30-minute prompt a prompt engineer writes

A serious prompt engineer goes further. They'll write something like:

You are an investment analyst applying Warren Buffett's documented
methodology. For every evaluation:

1. Start with Circle of Competence — can I explain this business
   in two sentences to a twelve-year-old? If no, stop.
2. Apply Mr. Market — assume the market price is an opinion,
   not a verdict. Price ≠ value.
3. Calculate Owner Earnings, not reported earnings: net income
   plus depreciation minus maintenance capex...

[~800 words of structured instructions continue]

When this is enough:

• You use this specific persona 50+ times over the next year
• You've read enough Buffett to know what the best prompts contain
• You enjoy prompt engineering as a skill worth developing
• You'll maintain this prompt as Claude's models evolve

What authority.md gives you

For $4.99 and 60 seconds, you receive:

• ~2,000 word engineered .md file with YAML frontmatter
• Five named mental models (not invented — referenced from Buffett's documented work)
• Decision heuristics he actually used, with source grounding
• Signature questions in his voice
• Anti-patterns — what he would not do (this is the hardest part to write yourself)
• A copy-paste activation prompt
• A worked example showing the framework applied
• Delivery as both a native Claude Skill (.zip) and plain .md for any LLM

The question isn't whether you could write this yourself. It's whether the time you'd spend is worth $4.99.

Honest cost comparison

Here's the rough time investment for each approach:

If your hourly value is over $10, writing a prompt engineered to 80% quality takes longer than the difference in output quality justifies. If your hourly value is over $30, the math doesn't require debate.

The quality gap nobody mentions

The part most DIY prompts miss entirely: anti-patterns.

It's easy to write "think like Warren Buffett — value, moats, long term." It's hard to write what Buffett would explicitly not do.

Buffett would NOT use leverage to amplify returns. "To make money they didn't have and didn't need, they risked what they did have and did need."

Buffett would NOT trade in and out of positions based on macro forecasts — nobody predicts the economy reliably, not even the people paid to.

Buffett would NOT invest in businesses he cannot explain to a twelve-year-old in two sentences.

These are the specific rejections that define the framework by negation. Most DIY prompts miss them entirely because most people haven't read enough of the source material to know them. We do this research for you. Every framework ships with anti-patterns.

When DIY genuinely wins

We'd tell you to skip the purchase if:

• You'll use it once. A 5-minute prompt is fine for one-off questions.
• You're a prompt engineer who enjoys the craft. You'll write something specific to your exact use case that's better than our general-purpose framework.
• You're testing whether persona prompts work for your workflow. Start DIY, confirm the value, upgrade when confirmed.
• Your framework is you. Your own decision-making heuristics are what you want the AI to use — not a famous person's. We don't sell that.

When $4.99 genuinely wins

Buy a framework when:

• You'll use it repeatedly. Once the Claude Skill is installed, every relevant query auto-invokes it. No re-prompting.
• You work with AI agents (Claude Code, custom tooling) and need structured instructions, not freeform prompts.
• Quality matters — this is for decisions, not entertainment.
• Research time is expensive. The five hours of reading we do per persona is reflected in the framework. You skip the reading.
• You want a reference. The .md file is a document you can share, review, and reason about — not a prompt buried in chat history.

The "months of engineering" claim

We say on the homepage: "We spent months engineering it properly." That's real. Here's what it covers for every framework:

1. Source research — primary documents (letters, books, interviews)
2. Mental model identification — not invented, extracted from documented thinking
3. Anti-pattern research — specific rejections, not generic advice
4. Voice calibration — the signature questions need to sound like the person
5. Technical engineering — Claude Skill frontmatter, auto-invocation triggers, compatibility testing
6. QA against banned platitudes (the generic-advice filter our generator checks)
7. Iterative refinement across hundreds of tests

If you're going to use this framework 20+ times over the next year, that work is worth $4.99.

Start here

• Browse frameworks — see what we've built. Hit "Compare" on any page to see our engineered framework vs a typical DIY prompt side-by-side, without paying.
• Read about the Claude Skill format if you're curious about the auto-invocation mechanism.
• Our activation prompt guide if you're building agents and want to see the technical integration.

Still on the fence? Try a flagship first. Warren Buffett, Kobe Bryant, Richard Feynman, Alex Hormozi — these are the most heavily engineered frameworks in our catalogue. If they don't feel meaningfully better than what you'd write yourself, reply to the delivery email and we refund within 24 hours.

---

Updated quarterly. Last reviewed: April 2026.