Some weeks ago I was working on a a proposal and a writer sent me a deck with early ideas. I opened the deck ready to feel inspired by honest thinking full of potential but then… I was smacked in the face by slop.
Everything in that deck was all one hundred percent produced in ChatGPT and shamelessly copy-pasted (em dashes and all). Then I read the email the deck was attached to and I noticed it was too long, too proper, too bland, and too empty so it seems I had been hit by shameless work slop twice. Or maybe not, maybe it was written by a LinkedIn heavy user but then I thought, what is the difference between a LinkedIn fan and ChatGPT? Does it matter? Should I get offended by AI-slop and be fine with human slop?
The value of quantity
Here is a list of brand platitudes I have collected over the years and that work like little clues to design-code oneself as a status brander:
Bold, inclusion, diversity, personal, belonging, for all, inviting, taking risks, challenge convention, leadership, simplicity, clarity, safety, meaning, experience, community, emotion, our practice, authentic, synergetic, physicality, human first, redefine, clarity, workshops, rethink, moving away from tools, stand out, tactile, experimentation, investigation, materialise, strategic, expression, prioritise, approach, compelling, awareness, purpose, research, solving problems, refine, avoid safe, expressive, serve, human, global, forefront, iconic, elevate, transformation, care, play, change, push, by design, natural, permeate, function, platform, change, humility, future, progressive, useful, for humans, break boundaries, innovative, looking back, contrarian bets, setting the cultural context, marketing vehicle, conviction.
I took this list of magic design words, fed them to ChatGPT and asked for a case study for a fictional coffee brand called “Drip Society” (name that I also chatgeepeteed by the way) and here is what I got in 3 seconds:
Drip Society, a Shoreditch café born in 2025, needed a brand identity that reflected its bold attitude and inclusive spirit. Our task was to design a system that stood out in a saturated market while staying human, personal, and authentic.
Our practice began with workshops to rethink what a café brand could represent. Rather than tools or trends, we focused on experience, community, and belonging. The research phase revealed a clear opportunity: most cafés chased minimalism, but few offered emotion or meaning. Drip Society would change that.
The wordmark was built around movement and flow, echoing the physicality of coffee itself. Its curved forms feel expressive yet simple. A palette of graphite, warm white, and copper conveys both clarity and depth. Every touchpoint, from packaging to signage, was treated as tactile and inviting.
Typography combined a grounded sans serif with a soft serif accent, balancing clarity and experimentation. This duality framed Drip Society as confident but never corporate, expressive without noise.
Strategically, the identity became a platform for inclusion, diversity, and care. It avoids safe aesthetics and instead embraces play, materiality, and awareness. Every choice aimed to elevate the everyday ritual of coffee into something purposeful and human first.
The result is a compelling, progressive identity that captures leadership through simplicity and meaning. By design, Drip Society redefines café branding for the future: bold, clear, and deeply human.
Don’t ask if it is good (because it is not) but ask if it is on par with 80% of case studies you have read online and yes, it fucking is. So what is the value quantity? Zero. End of article… or maybe not, there is more to say.
Automating the automation
I have been saving design articles in Notion since they launched so now I have a couple hundred articles that I fed to Notion’s AI and asked it to generate a new article using my style. Here is what I got in 5 seconds:
There is a point in every hype cycle where the product barely matters and the sentence does.
Not “brand story.” A single line the market can repeat without choking on it. “AI safety.” “Creator economy.” “Post‑naive internet.” These phrases arrive, and in a few weeks decks, roadmaps and budget lines quietly rearrange themselves around them.
Founders feel this first. You sit with someone building something genuinely odd, and watch the gravity of familiar language win. The architecture is duct tape and cron jobs, but the pitch wants to be “Cursor for lawyers” or “AI‑native vertical SaaS for logistics.” The phrase isn’t the truth. It’s a passport: proof you belong in a category slide.
Investors live inside these passports. They cannot live in every product, so they live in words. A new phrase that is legible and just risky enough is a gift: something to write in a subject line, something to say in a partner meeting without blushing.
That makes language the first interface. Before anyone touches your UI, they are touching your nouns. The interesting question is not “does this sound nice?” but “what does this sentence quietly commit us to?” Jason Wei’s “asymmetry of verification” is the right lens here. “We beat benchmark X by 5%” is heavy: there is a dashboard waiting for it. “We reinvent human connection” is weightless: it never has to hit anything solid.
Most contemporary positioning lives in that weightless zone. It has the right adjectives and the right temperature. It sounds like a safe average. Probabilistic prose: exactly what a large model would write if you prompted it with “Series A landing page, tasteful.”
The market does not move on averages. It moves on edges. The phrases that actually stick are slightly embarrassing at first. They name a pressure everyone feels but nobody has nailed yet. “Post‑naive internet” didn’t invent the exhaustion with platform utopianism; it gave the feeling a handle and a timestamp.
Designers are in a good place to work with this, if we stop treating language as a last‑minute layer. The job is not “pick a tone and massage the tagline.” It is to design words with the same attention we give to grids: constraints, edge‑cases, failure modes.
At the same time, if every line has to be a metric, you end up with a spec, not a story. The useful zone is the border: language that is just verifiable enough to be meaningful, but still strange enough to feel new. “AI for slides” is forgettable. “PowerPoint that argues back” is a picture. Only one of those will get repeated in a bar.
Underneath the slogans there is a quieter discipline: paying attention to what people actually steal from you. The best phrases leak from support tickets, early demos, off‑hand comments — moments where a room goes quiet for half a second because someone has said something obviously true and slightly unsettling. That is usability testing for language.
The ethical risk is speed. You can move language much faster than reality. A homepage update takes an afternoon; a datacenter, a team, a real capability take years. You can try to use words as leverage over reality — speak the future loudly enough and hope the rest catches up — but that is how you end up with beautiful decks and empty products.
A saner game is smaller and more patient. Treat words as a way of making reality sharper rather than shinier. Name the tension before you promise the resolution. Keep a short, deliberate vocabulary instead of a 40‑page “platform”: a handful of nouns and metaphors you are willing to see in other people’s slides three years from now.
In a world that is now literally scoreboard‑ing “narrative dominance,” that restraint is weirdly radical. Anyone can benchmark an LLM. Anyone can remix a meme. Finding ten heavy, precise strings for what you are actually doing — and refusing to let them dissolve into mush — is harder.
But language is just another material. You already know how to work with constraint and negative space. Do the same here. Treat every phrase you ship as a surface someone will grab before they ever touch your product. Make sure there is something solid underneath their hand.
Don’t ask if it is good, ask how good it is for 5 seconds and yes, it is fucking good for 5 seconds. So what is the value of quantity then? Less than zero.
Quantity is detrimental
There was a time when being a smooth talker who could spit out good sounding consultant non-sense had a lot of value but today is just human slop. The longer something is, the higher the chances it will be AI summarised and ignored.
And what is long? I say two paragraphs. More than two paragraphs will be suspected of being AI written.
Artisanal slop and automated slop is still slop
Something becomes slop once you understand the formula that produced it but you don’t need a machine to produce slop, you just need a formula.
B movies were artisanal slop, cheap operas were slop, chivalric stories were slop too and Don Quijote got sloped hard. Slop churned out by hand is still slop and has the same effects. Were the drawings in the Chauvet Cave neolithic slop? In their time they probably were.
But slop has a positive side, industrial slop tends to accidentally pay unique creators to do their thing so, for each river of slop you get a couple of David Lynches. But don’t get too carried away, even good art is also slop (most movies from Dario Argento are slop).
Everything is (mostly) slop. The only question worth asking is this one, does it matter if it is artisanal slop or automated slop?
Everybody pretends but we know
I think AI generation works like doom scrolling and dirty pictures. We pretend to read poetry in our free time but we know we doom scroll and look at dirty pictures. We pretend to hate slop in public but we consume slop and generate slop in private (and we know everybody else does the same).
So now that we have stablished that slop is old, that artisanal slop and automated slop are both slop, that mostly everything is slop, and that we are all guilty of consuming and producing slop, let’s go back to work slop.
Artisanal slop and automated slop
I was thinking all of this while sitting in front of that AI generated slop of a deck. I was dumbfounded but I had to do something. Should I ask an LLM to reword it? Should I do nothing? Should I slop it myself?
At the end I redid everything by hand and with good intentions like the idiot that I am. I reworded it, I reduced adjectives by 90%, I forced it to take a stance on something, and I did it all knowing the client will probably AI summarise it thinking I Chatgepiteed it because everybody else does.
Are you still here?
If you are human and still reading this, hi. What should we do my fellow bipedal consumer unit? Join the automated slop economy or rebel against it? Join the slop carrousel and prompt all of our outputs? Or pretend it is all fine and take a stand for artisanal slop? Or worse, pretend to take a stand and slop it in private?
I know that whatever we do, it will be fed as training data, summarised, monetised, and assumed to be slop so… I don’t know, do whatever, but for the love of God, at least pretend you half wrote it and don’t send me raw geepitee slop.