Full transcript
0:00Back in the early days of this channel,
0:02I spent a lot of time just doing videos
0:03about cool prompts that I was using and
0:07discovering for the creative writing
0:10like ecosystem. And I realized I haven't
0:13really done a whole lot of videos like
0:16that in a while. So, this is going to be
0:18the first of hopefully a series of
0:20videos just focused on some cool
0:23prompts. And in this one, I'm talking
0:25about what I'm calling the logic check
0:28prompt, which is just a prompt that I've
0:30developed to go through any particular
0:33story related element, particularly in
0:35the early planning stage of your books.
0:38Give it a logic and plausibility check.
0:40See like does this all make sense? Are
0:42there any sort of plot holes or things
0:45that just don't really make sense when
0:47you think about it going into that? No.
0:49I developed this through the process of
0:51creating an automation, but you can use
0:53this just by itself. You can create a
0:56project that specializes in running this
0:58prompt. You can just run it as a single
1:01chat thing. Uh or you can integrate it
1:04into whatever program you choose.
1:10Now, if we go here to my handydandy
1:15Jason's prompt document, which by the
1:17way, you can get for free by signing up
1:19for the waiting list for Story Hacker
1:20Gold link down below. You don't have to
1:22actually sign up for Story Hacker Gold.
1:24You just join the wait list. And if you
1:25do want to go and take it to the next
1:27level, you can, of course, join the gold
1:29when it opens up. But in the meantime,
1:31even if you want to join the Storyhacker
1:34Gold program, you can join the waiting
1:36list and you'll get this while you wait.
1:37All right. Uh, this prompt document with
1:40all of the prompts I've talked about so
1:42far to which I have recently added this
1:44logic check prompt. Now, the way checks
1:47work and the way I work with them is
1:49that we take a model, run a piece of
1:53content through it, and say check for
1:57these things one, two, three, etc. And
2:00then we have it run that check and that
2:02we then take that improvement plan and
2:04apply it to the original thing and are
2:06like okay now rewrite this to do these
2:10things that the improvement plan says
2:12but don't change anything else. And what
2:15that does is it allows us to
2:17incrementally improve on an original
2:19text. Whether that's an actual chapter
2:22of your book or whether it's a story
2:24dossier or a character bio or whatever
2:27that is. uh you can use a check like
2:29this to do that. So it's a two-step
2:31process. The second step is really
2:32simple. You just say here's the original
2:34thing. Here's my improvement plan.
2:36Please implement the change of the
2:38improvement plan and change nothing
2:39else. That's all you have to do. Let's
2:41talk about the logic check itself. I'll
2:43zoom in here a little bit so you can see
2:45it. Now I've designed this one. The one
2:47that's actually in my document is a
2:50logic check for a story dossier. So this
2:52is early in the writing process when
2:54we've kind of as we're assembling all of
2:57the different pieces of what we want in
3:00our book. So that's this is just for the
3:02story dossier. But if you want, you can
3:05just tweak this prompt so it works for a
3:08character bio, for a chapter plan, for
3:10you know an outline, whatever it is that
3:12you want. Just so you know, as I go
3:14through and read through the prompt,
3:15it'll we're talking just about the story
3:18dossier, which is part of my process,
3:21which you can see early on. So, it
3:23starts out something like this. You will
3:24be auditing an early stage story dossier
3:26for logical consistency and
3:28plausibility. Which, by the way, you can
3:29use this, even if you don't use AI for
3:32the actual writing of your book, that
3:33you can run your own brainstormed ideas
3:36and things like that, your plans, your
3:37outlines through a prompt like this to
3:39help you out. The dossier contains a
3:41premise, pitch, character list, and
3:43world building elements for a story
3:45concept. You would, of course, change
3:46that line. If you had something
3:48different from a dossier, here's the
3:50dossier to audit. And then you paste
3:53whatever your dossier looks like in
3:55here. Then we and and keep in mind, I've
3:58also got these XML tags. Um, I've talked
4:01about these a lot. This is just a good
4:02way for AI to really understand in neat
4:04little boxes where everything is, right?
4:07All right. And then here under
4:08instructions, if we say your task is to
4:10identify logical inconsistencies,
4:12implausible plot elements, character
4:13world mismatches, and worldbuilding
4:15conflicts that would undermine the
4:16story's internal coherence. Perform the
4:19following six audit checks. So, we got
4:21six of these. Premise logic check. Does
4:24the core premise hold together
4:25internally? Are the central conflict and
4:27stakes consistent with the stated world
4:29rules? Does the basic whatif or setup
4:31make logical sense? This is something I
4:33found is often a problem with some of
4:37the output that AI gives you. It gives
4:38you something that kind of sounds okay,
4:41but then when you really dig deep into
4:44it and think about it, you're like that
4:46actually like that character, it
4:47wouldn't really make sense for them to
4:49get that power at this time or like
4:51that's really convenient for the plot,
4:53all that stuff. There's often just
4:55slight logical or implausible
4:58misalignments there. All right. And
4:59number two is character world fit. Do
5:01the character roles, goals, and
5:02capabilities make sense given the world
5:04building? And does any character's
5:05existence contradict stated world rules?
5:07Do character relationships and
5:08motivations align with the premise and
5:10pitch? Are character power levels,
5:11social positions, or abilities plausible
5:13in this world? Number three is world
5:16building coherence. Do the world
5:17elements support and enable the premise?
5:19Are there conflicting world rules stated
5:21in different sections? Do geography,
5:23technology level, social structure, and
5:25magic/tech rules work together
5:27logically? Uh, you can obviously modify
5:29that if you're not writing fantasy or
5:31science fiction. Are there internal
5:32contradictions in how this world
5:34operates? Number four is plot setup
5:36plausibility given the pitch. Are there
5:38obvious logistical impossibilities?
5:40That's another one important for the AI
5:42to really think through. Are there
5:43character motivation gaps that would
5:44prevent the story from starting? Does
5:46the antagonist's power and position make
5:48sense relative to the protagonist
5:50starting point? Is the inciting incident
5:52plausible given the world and the
5:53characters? Number five is early stage
5:55convenience flags. This is a big one.
5:57Does the premise rely on unlikely
5:59coincidences without justification?
6:01That's something that AI does a lot. It
6:02just says this, you know, this sort of
6:04thing happens, but then you're like,
6:05well, why does that happen? Are there
6:07characters who exist purely to solve
6:09plot problems with no other narrative
6:11purpose? Does the protagonist have
6:12suspiciously perfect allies or resources
6:14that feel contrived? Are there because
6:16the plot needs it elements with no in
6:18world logic? This is a problem that both
6:20AI and most humans share when coming up
6:23with your plot. A lot of times where
6:24jurists would go like yeah but this
6:26needs to happen and you sort of force
6:28your characters down that path. It
6:30becomes a lot harder to figure out how
6:31can you do that naturally. So this is
6:33part of the prompt looks for that. And
6:35then number six is specific fixes. Uh
6:37for each major issue identified provide
6:40concrete suggestions to resolve the
6:42problem while preserving the core pitch
6:43and appeal of the story. For each of the
6:44issues you identify, use this format. Uh
6:47and then I have I give it some
6:49formatting things here. Here is an
6:51example of a well formatted issue that I
6:53put there. And then uh this is something
6:56that was suggested to me by the Claude
6:59prompt generator as I was trying to
7:01refine this prompt using that generator
7:03which is having it do this scratchpad
7:06thing. Before you write your final
7:07audit, use a scratch pad to
7:09systematically work through each section
7:10of the dossier. So it's providing this
7:13assuming that the model you're going to
7:14use is a thinking model that thinks
7:17through things before it actually
7:18creates the thing. First identify the
7:20core premise and main pitch. List all
7:21the characters in their stated roles,
7:23capabilities. List all world building
7:24rules and elements. Check each audit
7:26category systematically. Note any
7:27contradictions and plausibilities or
7:29logic gaps for each issue. Think through
7:31what specifically breaks and how it
7:32could be fixed. And then I have it say
7:34structure your final audit report with
7:36these six sections. 1 2 3 4 5 six that
7:38we already went over. If a section has
7:40no issues, state no significant issues
7:43identified in this category. In section
7:45six, specific fixes, provide three to
7:47six concrete, actionable suggestions
7:49that address the most critical problems
7:51while preserving the story's core
7:52appeal. Your final output should contain
7:54only the six section audit report, not
7:56the scratch pad. Each issue should be
7:57clearly formatted with the structure
7:59shown in the example above. And then we
8:00end the instructions. So this is the
8:02general idea behind the logic check. Now
8:05let's talk about what model you want to
8:07use with this. And first of all, you
8:08definitely want a thinking model because
8:10this is very much a thinking type of
8:12task. And while a lot of the checks that
8:14I do don't necessarily need to have
8:16really robust models, this is one that I
8:20think you do need a really smart and
8:23probably slightly more expensive model
8:25to run on your dossier or whatever it is
8:28you're running it on because it requires
8:30such a logical thought process. Like we
8:33really want it to think like a human as
8:35much as possible. And so to get the best
8:37results, you're going to want a model
8:39like currently as of today that I'm
8:42filming this, the best ones would
8:43probably be Gemini 3 or Claude Opus 4.6
8:47cuz they have some of the best reasoning
8:49capability out there right now. So I
8:51would use one of those. Obviously Claude
8:53Opus 4.6 is a little bit more of an
8:55expensive model. So you could use Gemini
8:573 if you want to save on cost. But if
9:00you're only running this on a single
9:02story dossier for your entire book, I
9:04don't have any problem using the most
9:06expensive model out there. And if you
9:07get cloud opus 4.6 through paid
9:10subscription to cloud or something like
9:11that, better yet, right? So that is the
9:14idea of the logic prompt. So what you
9:16would do is you'd run this on your story
9:17dossier. You would then run a second
9:20prompt and say, "Okay, now that you've
9:22created this these recommended fixes, go
9:24ahead and implement those fixes, but
9:26don't change anything else about the
9:27story dossier." and it would go and do
9:29that. And so once again, if you want to
9:31get your hands on this prompt, you can
9:33go ahead and join the wait list down
9:35below. Most days it is on a wait list,
9:38but you might see it like if you just
9:40happen to show up the day that the
9:42membership is open, you won't be able to
9:44see it as a wait list. So you if you
9:46just want the prompts and you don't want
9:47to actually join the group, you'd have
9:49to wait a day or so for the wait list to
9:52come back. That's just kind of like how
9:53it works. But anyway, go check out the
9:55link below to get this and I will see
9:57you in the next