dantepeppermint.com / field-notes
Thinking on brand, positioning, and the work.
Field Notes is thinking on brand, positioning, creative strategy, and original thought. Essays on what separates real creative work from the average. Written for people who build brands, make things, and care about doing it right. Each piece is a perspective, not instruction, not trend reporting. Just direct thinking about the work that matters.
Most people don't need a better ChatGPT. They need a different kind of AI entirely. The honest version of the question. What are you actually shopping for? , has four different right answers.
Claude for prose. Perplexity for research. Embedded tools for operational speed. And a smaller, opinionated category of thinking partners for the work that matters. Stack them. Don't replace.
Most prompt lists are recycled garbage. These thirty are pulled from real brand work. Positioning, messaging, naming, audience. They give the model something to work against, not just a job description.
They name the failure mode they're trying to avoid. And they ask for a kind of thinking, not a kind of output. Steal them, fork them, use them in any AI tool.
It's not a flaw. It's the design. ChatGPT sounds generic because it's optimized to sound generic. Predicting the most statistically likely next word, then sanded smooth by human raters trained to prefer the inoffensive.
The mechanism is the message. The em-dashes, the hedging, the symmetrical sentences. They're features of how the dominant tool works, not bugs to be patched. Here's how to stop it.
Most teams "train" an AI by uploading a style guide and three past campaigns. That is not training. That is reference material. And reference material, on its own, produces an average of what was referenced.
A brand voice is not a tone. It is a position. The fix is architectural, not editorial. Here is the difference between an AI that has read about your brand and an AI that thinks from inside it.
If your AI output sounds like everyone else's AI output, the model is not broken. It is doing exactly what it was built to do. Produce the most statistically likely next token. The most statistically likely answer is, by definition, the average. Averages do not have positions.
Three kinds of generic, three different fixes. And the test for whether you have actually escaped the average.
Marketing teams keep evaluating AI tools and getting disappointed. The tool was usually fine. The brief was the problem. And until teams start treating the brief as the variable instead of the tool, the next AI subscription will fail in the same way the last one did.
The four things a real brief contains: a position, a refusal list, the reader's prior, and the verdict the work has to deliver.
An honest comparison of the three frontier AI models for marketing strategy, brand positioning, and creative thinking. With a verdict on which to use when.
None of the three is best at everything. ChatGPT is fast and multimodal. Gemini is real-time and well-sourced. Claude pushes back. The differences matter more than the benchmarks.
Everyone at Coachella could listen to those artists for free. On Spotify, on YouTube, on any device they own. The music isn't the product. The lineup is. The act of choosing. Who gets in, who doesn't, in what order, on what stage. Is the thing people are actually paying for.
Taste is a brand position. The refusal to include everything is the promise. What you leave out is as load-bearing as what you put in.
The moment you start treating AI as a generation machine, you lose something important. The frame you bring to the tool shapes everything about what it can do with you. A generator runs a process. A collaborator thinks with you.
Collaboration requires you to remain genuinely open to the process changing where you end up. If you already know exactly what the output should be, you don't need a collaborator. You need a typist.
The most common complaint from people who use AI for creative work isn't that the output is bad. It's that the output doesn't sound like them. This isn't a flaw in the tool. It's a consequence of how the tool is typically used.
When you hand a creative problem to AI without sufficient constraints, it defaults to the average of what it's seen. The average is competent. It is not distinctive.
Most AI is designed to be opinionless. Not because the engineers forgot to give it opinions. But because opinions are expensive. They alienate users who disagree. They make AI harder to sell to the broadest possible market. Neutrality is the safer business decision.
The result is AI that's useful the way a dictionary is useful. Full. Accurate. Utterly without a point of view.
Brand positioning lives in a comfortable zone until something challenges it. Most positioning work never gets challenged. It gets approved, written into brand guidelines, and left alone until someone notices it isn't working.
Pressure testing is the practice of finding where your positioning fails before that happens. And an AI thinking partner is one of the most useful tools available for this. If you use it as a sparring partner rather than a validator.
There are two ways to talk to an AI. One produces generic output. The other produces something worth using. The difference isn't the AI. It's how you set up the conversation.
Most people interact with AI the way they fill out a form. Creative directors brief collaborators. That distinction determines everything about the output.
Artemis isn't the nostalgia mission. That's the easy read. America going back to the moon, retreading something Apollo proved fifty years ago. But the actual goal of Artemis isn't to revisit what happened. It's to build what comes next.
A lunar Gateway station. A permanent platform for reaching Mars. The moon isn't the destination. It's the foundation. And buried in that logic is a precise framework for how creative thinking actually works.
The infrastructure has won. Not deliberately, not through malice, but through sheer structural incentive. The platforms that distribute attention don't reward what breaks the pattern. They reward what fits it. And so the best minds in culture have internalized the logic: why spend energy on what won't travel?
This isn't about laziness or corporate compromise. This is about a system that has made originality economically irrational.
The Gen Z marketing brief has become its own genre of uselessness. Conference slides full of demographic stats. Agency decks that confidently explain what "they care about." Articles that tell you Gen Z wants brands to be "authentic," to "take a stand," to "meet them where they are."
All of it true in a way that's completely useless. True like "make good food" is true for a restaurant. Not wrong. Just not a strategy.
There's a category of brand that doesn't behave like a brand. It doesn't chase you with retargeting ads. It doesn't remind you that you haven't used it in three days. It doesn't send a survey asking how it can improve.
It doesn't need to. Because you already reached for it before your eyes fully opened. That's a different kind of loyalty. And most brands can't buy their way into it.
The comfortable thing is almost never the right thing. Not in creative work. The comfortable thing is what you already know how to do, already know how to say, already know how to build. The comfortable thing is the version of the work that you could finish in your sleep. And the problem is that it will feel like you did.
There's a reason the best work tends to come from the edge of what someone knows. Not from the center of their expertise. Not from the place they're most certain.
Most of what gets called intelligence is pattern-matching at scale. Match enough patterns, in enough contexts, and you can produce outputs that look like thinking. They have the shape of thought. They follow the grammar of reasoning.
But there's something missing in the center. A perspective, a point of view, a position on the question that doesn't change based on what you want to hear.
Most prompting advice is about volume. Write longer prompts. Add context. Be more specific. These things help. They are also not the thing. The real variable isn't length. It's form.
Questions invite answers. Answers don't require reasoning. They require selection. The most effective way to use an AI thinking partner isn't to ask it things. It's to give it a situation.
Here's a thing that happens when you ask most AI tools for creative help: you get something that sounds right. The sentences are grammatically correct. The structure is familiar. The advice lands in a register you've heard before. Confident, slightly hedged, covering the main angles.
And when you read it, you feel vaguely worse about your work. Not because it's wrong. Because it's average. It reflects the mean of everything that's ever been written on the subject. And the mean is not where the interesting work lives.
Good positioning feels wrong at first. Not wrong in the sense of incorrect. Wrong in the sense of uncomfortable. It creates friction. It doesn't play it safe. And the moment you hear universal agreement in a room, you should get suspicious that you've watered it down into irrelevance.
The consensus trap is real. A committee destroys positioning not through malice but through reasonableness. Every objection gets addressed. Every rough edge gets smoothed. Until what's left is something that offends no one and moves nothing.
The problem with using most AI for brand positioning is this: AI is trained to say what sounds correct. It reads your brief and optimizes for the most plausible answer. Not the most honest one. That's a different problem than brand positioning actually requires.
Dante approaches it differently. A thinking partner doesn't validate your brief. It interrogates it.
A chatbot is optimized for satisfaction. Its entire architecture. The training, the fine-tuning, the reinforcement from human feedback. Is pointed at one thing: giving you something you find acceptable. It wants to resolve your question.
That is not what a thought does. A thought doesn't resolve. It opens. It complicates. It introduces tension where you expected clarity. Because that's what real thinking actually feels like.
We're asking AI to do the thing it's worst at: make bets. And we've dressed it up as strategy. A leader sits down with Claude or ChatGPT and asks: "What should our brand positioning be?" The AI returns five positioning statements, each backed by market research, competitive analysis, and consumer sentiment. Clean. Full. Wrong.
Not factually wrong. Directionally wrong. The kind of wrong that takes eighteen months to discover.
Most brand systems fail because they document decisions instead of defending them. They answer the wrong question: what did we decide? Instead of the right one: why does any of this matter?
This one starts from a different place. A visual system should have an opinion. Not just a palette. The decisions here weren't made by committee. They were made to mean something.