Ordering Coffee from an Algorithm
AI isn’t just teaching machines to think. It’s teaching humans to speak clearly again.
We built algorithms to talk like us, and now they’re teaching us how to listen again.
I think people order coffee as if they’re finding the right permutation and combination to re-create the Powerpuff Girls. Oat-milk cappuccino, half-shot, extra foam, not too hot, not too cold - something that says I’m emotionally stable but still edgy.
The barista nods gravely, presses a few buttons, and delivers a drink that tastes like regret and overthinking - or poonai susu (cat piss), in my mother’s parlance.
Talking to AI feels like ordering coffee at a pretentious café: too many options, too much jargon, and the quiet terror of ordering something you don’t want. We don’t know what we want, exactly; we just know we want to sound intelligent while asking for it. So we over-describe, over-qualify, and then feel betrayed when the output comes back flat. Feed a model confusion and it serves confusion - perfect punctuation, excellent grammar, and slop content. Garbage stays garbage, even when you pronounce it gahr-baa-dʒ.
Prompting, in theory, is instruction. In practice, it has become performance. People trade “prompt hacks” on Reddit the way earlier generations swapped recipes. The goal is to appear witty, worldly, and slightly above sounding AI-like. The result often reads like a word salad that has anchovies or pizza with pineapples - yummy to some and 🤢 to others.
Comedian Biswa Kalyan Rath’s bit about ordering coffee is really about anxiety - the fear of getting something wrong in public; not just diffusion in liquids and gases. AI has become the same ritual. Everyone’s told to “use it,” but no one wants to look unsophisticated doing it. So we ask the machine to fix what we can’t articulate, and when it fails, we sigh like disappointed Indian parents. We don’t know what we expected, only that it wasn’t this.
The Accidental Barista
In my twenties, I spent (way) too much time around electrical and mechanical engineers who moonlighted as prog-rock musicians. They’d do engineering by day and perform Dream Theater by night. I was the odd one out: a sentence engineer among the mathematically inclined; rewriting their posters so they didn’t sound like a lab manual or look like a ransom note.
Once, I wrote a feature about their band, Black Earth. I thought it was immaculate: balanced structure, elegant subheads, not a comma astray. My editor said, “Good, but it lacks life. Spend more time with the band.” I did, and the rewrite worked, and somewhere along the way I thought the guitarist was really cute.
Aside: Vikram Sekar (guitars, backing vocals, and Viks Newsletter) and Subbu (drums and Chip Log) played in this band!
At the time, I thought “life” was code for sentimentality. Years later, I realized it was code for empathy - the ability to write so someone else can feel what you meant, because you can be precise and still fail to connect.
AI is exposing the same gap; it’s revealing how much of our communication depended on tone, context, and shared patience. When you strip all that away, what’s left is syntax, and syntax without empathy is sterile. Machines do syntax without empathy, so they can’t guess what you meant, so they give you what you said.
The Deaf Ear of the Machine
Humans aren’t so different from algorithms. We predict, we interpolate, and we fill gaps. Anyone with partial hearing knows conversation isn’t a clean stream of words; it’s reconstruction of what you heard, context, and facial cues and infer the rest. Communication only works because someone else extends the courtesy of understanding what you meant to say.
Machines don’t grant that grace. They mirror you back with merciless literal-ism. They over-understand, and that’s why AI feels uncanny: it reflects our half-formed thoughts with perfect grammar. For the first time, we are able to see exactly how unclear we sound.
Maybe that’s the real discomfort. The algorithm is the first conversational partner that refuses to indulge us - well when you tell it to get past the BS and be a sparring partner. When you ask it to “write a professional yet friendly note,” it faithfully reproduces your internal contradiction. It can proficiently format your social anxiety.
The Clarity Problem
We keep saying AI will teach machines to think, but the machines are also teaching us to mean what we say.
When you phrase something with care, anticipating how it might be read, you’re not just being polite; you’re practicing cognition. Clear expression assumes your listener matters. AI simply exposes how rarely we extend that courtesy.
Perhaps that’s why people call the technology “cold.” It’s impartial instead of being sentimental. It forces us to confront how much of our warmth lives in the spaces between words. Machines can fix your grammar, but not your grace. They can polish your syntax, but not your sincerity.
A Small Return to Meaning
Every major communication tool rewired what clarity meant. The printing press rewards structure. Google rewards keywords. AI rewards clarity - the ability to write so another can make sense of your ideas.
That may be its most useful feature. It’s making visible the murky soup we used to call communication: the filler emails, the vague memos, the passive-aggressive texts. AI cannot replace us - it hands our language back, stripped of its social camouflage - by asking: Is this really what you meant?
Maybe this is what progress sounds like: slightly awkward, occasionally repetitive, but finally aware of itself. We built algorithms to talk like us. They’re just returning the favor by making us listen to ourselves.
Thank you!