Full transcript
0:00When it comes to Chat GPT, just using it
0:02as your go-to tool for writing, trust
0:04me, I did this for way too long, means
0:06you're leaving massive quality and
0:08productivity on the table. So, in this
0:09video, I'm giving you the complete
0:11breakdown of why I actually quit using
0:13Chat GPT as my main tool, and what I
0:16switched to instead. We're going to go
0:17through the specific tools I now use
0:20every day, how they work together as a
0:22system, and why this shift completely
0:24changed the way I work. By the end,
0:26you'll know exactly whether this kind of
0:28setup makes sense for you as well. Okay,
0:30so why did I walk away from Chat GPT as
0:33my default? Look, it's not that Chat GPT
0:35is that bad. It will do in a pinch. But,
0:37most of what it offers is available in
0:39other tools and then some. So, let's go
0:42through why I unsubscribed from Chat GPT
0:44over a year ago, and why I've never gone
0:47back, which is saying something because
0:49I'm an influencer in this space, so I
0:50should have access to a lot of different
0:52tools, but I genuinely have never needed
0:55to use Chat GPT since I unsubscribed a
0:58year ago. Reason number one, the models
1:00aren't actually that good at writing.
1:03Now, this has been historically the
1:04case. We all know that OpenAI was the
1:06first one to really explode into the
1:09public consciousness when it comes to
1:11AI, even though AI had existed in
1:13various forms for a long time. It was
1:15Chat GPT that really set the bar
1:18initially. And so, because of that, a
1:19lot of people started out with Chat GPT
1:21and just assumed that that's, you know,
1:23the one that they would continue using.
1:25But, we quickly found out, even early
1:27on, that other models, such as Claude,
1:30were far more proficient at writing than
1:33any of the GPT models. This is true all
1:35the way back to GPT-4 and Claude 2,
1:38which now, looking back on even Claude
1:402, the amount of progress that we've
1:43made since then has been incredible.
1:45But, Claude 2 was definitely better at
1:47writing than GPT-4. And that is 100% the
1:50same today. If you look at the Claude
1:52models, if you look at a lot of
1:54different models, even Grok and some of
1:56those others, the writing quality tends
1:58to be a little bit better than the
2:00models coming from Chat GPT. I'll give
2:02you one example here. This is a short
2:05fantasy scene written with GPT 5.5 and
2:08another model called Aeon 2.0, which is
2:12a extremely inexpensive model. If we
2:14actually look at the pricing for these,
2:16GPT 5.5 is $5 per million tokens of
2:20input and $30 per million tokens of
2:21output, whereas Aeon 2 is only 80 cents
2:25for the input or and $1.60 for output,
2:28which means you could write a novel
2:30that's almost 30 times the size or at
2:33least like 20 times the size using Aeon
2:352 and it would be the same cost as what
2:40GPT 5.5 would give you. So, literally
2:43we're looking at like 20 books instead
2:45of one. But, let's actually look at the
2:46pros itself. So, this is what GPT 5.5
2:49wrote. "Lathan did not wait to be
2:50announced. The bronze doors of the
2:52throne room boomed against the inner
2:53walls as he shouldered through them. The
2:55sound rolling up into the vaulted dark
2:57like thunder trapped beneath the
2:58mountain." Lot of metaphors already.
3:00"Every head turned. Courtiers in wolf
3:02fur mantles stiffened along the
3:04blackstone pillars. A pair of hearth
3:06servants froze with armfuls of split
3:08pine. At the far end of the hall beneath
3:10the great antlered crown of Varrolden,
3:12his father sat unmoved upon the throne.
3:14King Stor looked as though the mountain
3:16had carved him and forgotten to give him
3:18blood." Another weird metaphor there
3:19that doesn't make too much sense.
3:21"Father, Lathan said, voice raw from
3:23shouting at the gate, from shouting too
3:24late. A man died outside our walls
3:26tonight." The hall seemed to draw one
3:28long breath. All right, so clearly very
3:30wordy and we're kind of in a very
3:33narrative distance from the character.
3:35We're sort of just being described
3:36what's happening. Let's look at what
3:38Aeon 2.0 gave us. "The stone corridors
3:40of Varrolden's royal keep had never felt
3:42narrower than they did now." So, already
3:45we're getting into the feelings of the
3:46character. "Pressing in on Lathan as he
3:48strode through the throne room. The
3:50scream still echoed in his skull, that
3:51desperate raw sound of a man being torn
3:53apart while the city watched, while they
3:55watched, while he watched, held back by
3:57hands that had no right to touch him.
3:58The guards at the door saw his face and
4:00stepped aside without a word. Good, he
4:01didn't have words for them anyway. The
4:03throne room opened before him, cavernous
4:05and cold, lit by a pale glow of lumen
4:07stones set into into the walls. His
4:09father sat upon the seat of Arlden, a
4:11simple thing of dark granite worn smooth
4:12by centuries of kings. Storr's face was
4:14weathered, his beard shot through with
4:16gray, his eyes carrying the weight of
4:17too many winters. "Father," Laethan's
4:19voice cracked through the chamber like a
4:20whip. He didn't bow. "Laethan," Storr's
4:23tone was measured, weary. "I take it
4:24you're here about the gate." So, already
4:26we can tell this is head and shoulders
4:29better than what GPT was giving us.
4:31Like, this is an independent model. This
4:32is a fine-tuned version of Deep Sea.
4:35Now, is it perfect? Of course not.
4:37There's still a few AI-isms in there.
4:39There's, you know, we've got an M dash
4:41here. I think it still kind of dwells
4:43too much on metaphor and description in
4:47certain things. But, you're clearly
4:48closer in terms of narrative distance to
4:51the character's mind, or sort of getting
4:53into his head and seeing how he thinks.
4:55I didn't get into the dialogue too much
4:56more, but Aean is really good at
4:58dialogue. So, this is just an example of
5:01how you can actually get far better
5:03writing out of other models than you can
5:06in ChatGPT. Reason number two, most
5:08other chatbots have the same abilities,
5:10but with better models. And the two that
5:12I use most frequently are Perplexity and
5:15Claude. Perplexity started out as more
5:18of a Google search replacement and got
5:21really, really good at that one thing.
5:23And I don't use it for everything, which
5:24is why I do also carry a Claude
5:26subscription. In addition to the Claude
5:28models generally being better than the
5:30GPT models, I also find that the tool
5:33itself, so the actual chat interface and
5:36the way it thinks through problems and
5:38things, which is not just something that
5:41depends on the models, but actually
5:42depends on how they've sort of coded up
5:44the software, tends to be more thorough,
5:46tends to be more accurate, has uh fewer
5:49issues there. And one of the nice things
5:51about Perplexity, that even the Claude
5:53subscription doesn't have, is Perplexity
5:55allows you to have multiple models. So,
5:58if I go into my Perplexity here, uh
6:00first of all, there's all kinds of
6:02different things that you can do like
6:04deep research and a model council. Uh
6:07some of these ChatGPT has, some it
6:08doesn't. Uh you can also connect it to a
6:10whole bunch of things, and Claude is the
6:12same as well. But, the thing I like
6:14about Perplexity most is that I can
6:16change models, and they've got their own
6:17proprietary models, they've got uh GPT
6:205.4. On the Max plan, you can get 5.5,
6:23but I don't have that. You also get
6:24Gemini, you get Claude, you get Kimmi,
6:26you get this model from Nvidia. There's
6:28more flexibility. Not a ton of
6:30flexibility, but it's just nice to be
6:32able to say, "You know what? I think
6:33this kind of problem is a better problem
6:36for Claude or for Kimmi K2 or something
6:39like that." Or maybe you want to get a
6:40little bit more usage out of the model,
6:42so you switch to a more inexpensive
6:44model. That sort of thing. Very nice to
6:47not be limited just to the GPT models.
6:50And for that matter, if you're using
6:52Claude, it's nice to not be limited to
6:54the Claude models as well. I like having
6:56variety, and so that's why I still keep
6:58Perplexity around, even though I do
6:59think that sometimes the Claude uh
7:02subscription tends to be a better
7:04powerhouse when it comes to actually
7:05getting work done. But, when it comes to
7:07just like, you know, making a search
7:09query or any sort of random research
7:11that I need to do, you know, various
7:13things, I like to go to Perplexity for
7:15all of that. All right, reason number
7:16three. If you want to write, you need to
7:18actually have a writing-specific AI
7:20tool, and not just a chatbot. Chatbots
7:23are great, they're incredibly simple and
7:25versatile. Because they're so simple,
7:26they are so versatile. But, they're not
7:28really designed with writing in mind,
7:31especially creative writing. There's all
7:33kinds of extra consideration that you
7:35want to have in there, like making sure
7:37you have a a dossier of your whole
7:40story, making sure you have your list of
7:42characters all fleshed out and that
7:43they're pulled into the story at the
7:44correct time so that the AI knows about
7:47like which characters are in the scene
7:48and all that. There's this whole
7:49workflow that a chatbot's just not super
7:52efficient at, which is why I always
7:54recommend even if you have a chatbot
7:56already that you also have a writing
7:59specific tool. Now, there are three that
8:01I've talked about a lot on this channel.
8:03They are Raptor Write, Novel Crafter,
8:05and Pseudo Write. And I've used all
8:07three, they're all good, they will all
8:09get the job done. And I haven't actually
8:11talked about them a lot recently because
8:13for one, I've been developing my own
8:15tool, which was just released, and that
8:18kind of makes it so I can't actually
8:20100% objectively talk about other tools,
8:22so I've just sort of stopped talking
8:24about a lot of them. And the other
8:25reason is that even before I developed
8:27my own tool, I'd actually quit all three
8:29of those in favor of automations, which
8:32is kind of like creating your own tool
8:34for your own purposes that can be
8:35writing specific or any other type of
8:39specific. And that brings me to reason
8:41number four, which is that I kind of
8:42abandoned Chat GPT and a lot of these
8:45writing specific tools for automations.
8:48At least when it comes to the actual
8:50work I was creating with AI, the sort of
8:53the powerhouse elements of AI. Now, I've
8:55done various videos on this channel
8:57about using a tool called N8N, which is
8:59my preferred automation tool. It does
9:02have a fairly high learning curve to get
9:04started, and I have a whole series of
9:05videos about that, but you can have
9:07automations that go anywhere from a
9:09simple automation like this one, which
9:12all this does is it helps you create a
9:15style guide that you can then
9:16incorporate to make AI sound more like
9:19you, all the way over to a behemoth like
9:21this one, which has like, I don't know
9:23how many steps, which is tailored to a
9:25specific task that I do in my business.
9:27And you can have automations like this
9:29that are tailored to your writing, you
9:30can have them tailored to marketing, to
9:32helping you create marketing materials,
9:34you can have them tailored to managing
9:35your email. There's all kinds of
9:37different things that you can use an
9:39automation tool like this for. Now, like
9:41I said, this does have a pretty strong
9:43learning curve, which is why I took most
9:45of the writing automations that I've
9:47developed and included them in my own
9:49StoryHackers OS tool because that
9:51allowed it to be a lot more accessible
9:53to a lot of people. But, I still
9:55recommend that people learn an
9:56automation platform like n8n. Another
9:58one that's maybe a little simpler to
10:00use, but not as powerful, is called
10:02Make. That one's worth checking out just
10:04because I feel like there are a lot of
10:05tasks that automation could take off of
10:08your plate so that you don't have to
10:10worry about those things and you can
10:11instead spend more of your time on those
10:13things that matter most to you. Now, if
10:15you want to know more about the tools
10:16that I do use and how I use them, I
10:18created the ultimate guide to go from
10:20zero to book in hand, which has
10:22everything you need to know about
10:23writing with AI, including the tools
10:24that I actively use today. So, go check
10:26out that video right there and I will
10:28see you in the next one.