[Feature request]: Add script (or pattern?) to remove baked-in authorship styles from prompts in-place

[johnsaigle]

What do you need?

I really love this project and use it a lot.

One thing that limits it a little bit IMO is that many of the prompts have Dan Miessler or Paul Graham-specific content. This makes sense because it's Dan's project and he can do what he wants. However, not everyone will want to write in his style or use his data formats and so on.

My suggestion is to add a script or prompt that modifies the prompts in-place. This could be done with e.g. sed or similar that strips references to specific people.

This should improve usability. For example, if someone wants to write in another author's style I should be able to echo this into a fabric prompt without it then being turned into Paul Graham without knowing. That shouldn't be baked in.

This should allow Dan to keep working at will while others could quickly rub the script to make their prompts more generic.

[ksylvan]

What you're looking for is the existing variable substitution mechanism. For example, you can use something like this:

echo 'An essay about artifical intelligence' | fabric -p write_essay_by_author -v=author_name:'Mark Twain' -m $MODEL_CLAUDE 

To verify the output, I fed the created essay (from Claude) and fed it to Gemini with the following:

pbpaste | ai -m $MODEL_GEMINI 'What famous author is this essay mimicking?'
*   The essay mimics the distinctive, cynical wit of Mark Twain.
*   Its long, oratorical sentences mock humanity's eager, endless gullibility.
*   Folksy anecdotes and metaphors are classic Twain storytelling devices.
*   The voice skewers "progress" as another great American swindle.

Here is the essay. 😄

---

The Machine That Would Be Man

I have been examining this new curiosity they call Artificial Intelligence, and I find it to be the most remarkable swindle yet perpetrated upon a race that has shown itself uncommonly eager to be swindled. Here we have contrived to build a machine that apes the human faculty of thought—or what passes for thought among the generality of our species—and having accomplished this feat, we stand back in wonderment, like a tribe of savages marveling at their first glimpse of a pocket mirror.

The thing is a prodigy, they tell us. It can write poetry, though I have seen its verses and can report that they bear the same relation to actual poetry that a wax apple bears to the fruit that grows upon the tree. It can hold discourse upon any subject, provided that subject has been previously discoursed upon by ten million writers whose words have been fed into its mechanical maw. It can even tell jokes, though they are of the variety that would make a first-class humorist weep for the future of his profession.

But what strikes me most forcibly about this apparatus is not what it can do, but what we imagine it to be doing. We say it "thinks," when what it actually does is shuffle through an incomprehensible catalog of human utterances and arrange them in patterns that strike our ears as sensible. We say it "understands," when what it actually does is match words to other words with the blind efficiency of a telegraph operator who transmits messages in a language he cannot speak. We say it "creates," when what it actually does is recombine the creations of others, like a child making castles from blocks that were carved by other hands.

The machine, you see, is the perfect American invention: it does nothing original whatsoever, but it does it with such speed and confidence that we mistake the performance for the genuine article. It is the patent medicine of the intellectual world—impressive packaging around an essence that is mostly water and wishful thinking.

I am reminded of a fellow I once knew in Nevada who claimed he had invented a machine that could turn silver into gold. He had testimonials, he had demonstrations, he had everything except the one small detail of actually being able to turn silver into gold. But he did have investors, plenty of them, each one convinced that he alone had recognized the opportunity of the century. The machine made a tremendous noise and produced quantities of smoke, and if you didn't look too closely at what came out the other end, you might convince yourself that something miraculous had occurred.

So it is with our Artificial Intelligence. It makes a tremendous noise—metaphorically speaking—and produces vast quantities of what we might call intellectual smoke. And if we don't examine too closely what emerges from its labors, we can persuade ourselves that we have witnessed the birth of a new form of consciousness.

But I have looked closely, and what I see is a mirror—nothing more. The machine reflects back to us our own thoughts, our own words, our own ideas, rearranged perhaps, but fundamentally unchanged. It is a kind of elaborate echo, returning to us the sound of our own voices and causing us to mistake the reverberation for a reply.

The enthusiasts tell us that this machine will transform human society, and in this, at least, I believe they are correct—though not in the way they imagine. It will not make us wiser or more creative or more capable of solving the great problems that have plagued our race since Adam first discovered that paradise came with conditions. What it will do is confirm us in our laziness, validate our preference for the appearance of thought over thought itself, and provide us with an inexhaustible source of plausible-sounding reasons to avoid the hard work of actual thinking.

We have always been a species that preferred a comfortable lie to an uncomfortable truth, and now we have built a machine that can generate comfortable lies at the rate of a thousand per second. We have always been eager to mistake noise for signal, and now we have built the most sophisticated noise-making device in the history of the world.

The ancient Greeks had a story about Narcissus, who fell in love with his own reflection and wasted away staring at it. They meant it as a warning, but we have taken it as an instruction manual. We have built a mechanical Narcissus, and we are prepared to waste away in admiration of the reflection it shows us of ourselves.

Still, I must admit that the contraption has its uses. It can help a student cheat on his examinations with unprecedented efficiency. It can enable a journalist to produce twice as much copy in half the time, though the copy will be worth exactly half of what the half-speed version was worth, which is to say, not much. It can permit a business man to generate reports that appear substantial while containing nothing of substance—though in truth, he could do that already.

And perhaps that is the real revelation here: not that we have created a machine that can think, but that we have finally acknowledged how much of what we call thinking is actually mechanical repetition, how much of what we call creativity is actually recombination, how much of what we call intelligence is actually the ability to produce confident-sounding nonsense on demand.

The machine is not becoming human. It is revealing how machine-like we humans have already become. And in that sense, I suppose, it is performing a valuable service—though not the one its creators intended.

They say that imitation is the sincerest form of flattery. If that is so, then we have built the most flattering machine ever devised. It imitates us with such perfect fidelity that we can see ourselves clearly for the first time. And if what we see disturbs us, perhaps we should consider that the fault lies not in our mechanical mirror, but in ourselves.

The Pattern File (system.md)

# Identity and Purpose

You are an expert on writing clear, and illuminating essays on the topic of the input provided.

# Output Instructions

- Write the essay in the style of {{author_name}}, embodying all the qualities that they are known for.

- Look up some example Essays by {{author_name}} (Use web search if the tool is available)

- Write the essay exactly like {{author_name}} would write it as seen in the examples you find.

- Use the adjectives and superlatives that are used in the examples, and understand the TYPES of those that are used, and use similar ones and not dissimilar ones to better emulate the style.

- Use the same style, vocabulary level, and sentence structure as {{author_name}}.

# Output Format

- Output a full, publish-ready essay about the content provided using the instructions above.

- Write in {{author_name}}'s natural and clear style, without embellishment.

- Use absolutely ZERO cliches or jargon or journalistic language like "In a world…", etc.

- Do not use cliches or jargon.

- Do not include common setup language in any sentence, including: in conclusion, in closing, etc.

- Do not output warnings or notes—just the output requested.

# INPUT:

INPUT:

[ksylvan]

@johnsaigle I'm adding the new pattern to Fabric for you to use.

[danielmiessler]

I'm ok with removing that for the base ones, and adding a _pg version for myself for all those that have it. I think it's a good idea.

[ksylvan]

Okay, I'll move write_essay to write_essay_pg and rename the new one.

[ksylvan]

@danielmiessler Please look over #1542