AI Knows Everything. It Just Needs Taste.

Why more options have become less useful — and why taste, context, and judgment are the real luxuries now.

Second Thoughts is Dandelion Chandelier’s ongoing series of sharp, slightly contrarian essays on luxury, style, taste, and desire — where hype, status, and cultural performance get a more intelligent second look.

AI has made answers abundant. It has not made judgment easier.

Modern search, social media, and generative AI are very good at giving us more: more recommendations, more rankings, more lists, more tabs, more confident summaries of things we may or may not actually want. What they are less good at is taste. Vale, Dandelion Chandelier’s AI taste assistant, was built for the moments when the answer depends on context, timing, social fluency, and the room: what to wear, where to go, what to see, what to gift, and what to skip. The next luxury is not more information. It is a better edit.

At a glance: AI taste assistant • too many options • better recommendations • taste and judgment • what to wear • where to go • what to skip

All photographs are original images by Pamela Thomas-Graham.

the age of too much

We used to treat access as the problem.

Access to the right restaurant. The right hotel. The right dress. The right exhibition. The right gift. The right book. The right person with the right answer.

Now access is everywhere. The problem has moved.

We do not suffer from not knowing enough. We suffer from being asked to sort too much, too fast, in public, with very little help from tools that mistake popularity for judgment.

That is why so many modern recommendations feel strangely inadequate. They answer the visible question and miss the real one.

Where should I go?

What should I wear?

What should I bring?

What is worth my time?

What should I skip?

These are not merely search questions. They are taste questions.

And taste requires context.

the midtown lunch problem

Last week, I needed to choose the right restaurant in Midtown for an upcoming business lunch.

Not just any business lunch. This was a lunch with someone one of my corporate boards was considering making an offer to join. I had been deputized to meet the person and probe for everything that would never appear on a résumé: how they presented themselves, their demeanor, their sense of humor, their ability to tell a story — all the things people mean when they talk about whether someone is “a good fit” for a board.

That meant the restaurant had to do several things at once. It needed some ambient energy, but it could not be too loud. The menu should have food that was good but not fussy, familiar but not boring, because the meal was not really about the food. The restaurant had to be close to my office, and close enough to theirs to feel considerate rather than inconvenient. It needed to feel polished, current, and comfortable, without becoming so sceney that the room itself took over the conversation.

I know exactly how wrong this can go because I have been on the other side of the table. Years ago, when I was being interviewed for a board seat myself, the person took me to a sushi restaurant. Which meant half of my mental energy went to using the chopsticks properly instead of engaging in the substance of the conversation.

It was a small thing, but not a small thing.

The room, the menu, the mechanics of the meal — all of it either supports the conversation or quietly sabotages it.

In other words, I was not looking for a popular answer.

I was looking for the right one for this very particular lunch.

why more answers are not the same as a better answer

Once upon a time, that would have meant a Google search. Now it means Google, TikTok, Instagram, ChatGPT, Gemini, Perplexity, and the rest of the modern parade.

Each will give you names. Many of those names will be good. Some will be sponsored. Some will be repeated so often they begin to sound authoritative by sheer force of frequency.

What none of them reliably gives you is the thing you actually need.

They do not tell you what the room feels like. Often, they do not provide an accurate read on whether the noise level leaves room for an actual conversation. Sometimes, they do not tell you whether the service is smooth enough to disappear into the background. They do not tell you whether the place signals confidence, discretion, and grown-up ease, or whether it is simply fashionable in a way that makes the whole exercise harder than it needs to be.

A recommendation without context is just another errand.

And that, to me, is the whole problem.

the new scarcity is judgment

We have made information so abundant that judgment has become the scarce resource.

Search is very good at surfacing options. AI is very good at making those options sound coherent. Neither, on its own, solves the harder problem: which one deserves your time, your money, your body, your attention, your evening, your reputation, your mood.

That distinction matters more than people admit. We still talk as though access to information is the great modern prize, when in reality the hard part is what comes after. The hard part is choosing. The hard part is sorting signal from noise, popularity from quality, and actual discernment from whatever happens to be trending among strangers with ring lights.

The modern internet is an overstocked department store of recommendations: many sponsored, many recycled, and most of them oddly unconcerned with whether the answer is right for you. It is abundance without hierarchy. Volume without judgment. A thousand names and very little shape.

For some categories of life, that is merely annoying. For others, it is exhausting.

The decisions that shape a day, a trip, a dinner, a meeting, a weekend, a room, a look, a gift, a social impression — those are not improved by being handed forty-seven tabs and a prayer. They are improved by context. By subtext. By someone, or something, that understands that the right answer is rarely the most popular one.

It is the one that fits the moment actually being lived.

Taste begins where abundance becomes a problem.

the best recommendation is still edited

This is why, when something really matters, I still go totally old school. I ask a friend.

Not just any friend, obviously. A friend who shares my taste level. One who is out and about enough to know what is new, what is over, what still has integrity, and what has somehow become impossible to hear yourself think in. A friend who understands that there is a meaningful difference between somewhere fashionable and somewhere right.

That is how I found 425 Park, one of the best newer places for a business lunch in Manhattan right now.

A good friend can give you in thirty seconds what the internet still cannot give you in thirty minutes: context, subtext, and a reason. Go here, not there. This room is better for a real conversation. That one is trying too hard. This one is elegant without advertising the fact. That one is all noise and expensive banquettes. This one still remembers what discretion is.

That, to me, is the gold standard.

But excellent friends, however useful, are not infrastructure. They are traveling, in meetings, or asleep in another time zone. They are sick, busy, offline, unreachable, or simply — how dare they? — living their own lives.

And then what?

culture is not a checklist

Restaurants are only the beginning.

I am an avid consumer of culture, and living in New York means I could spend all day every day in a museum, gallery, concert hall, or theater. Which would be fantastic. It is also impossible.

So the question is never whether there is enough to see. The question is what is actually worth making time for. It is easy enough for someone to say: go to The Metropolitan Museum of Art, go to MoMA, go to the Whitney. And of course sometimes the obvious answer is obvious for a reason.

What AI and search are worse at finding is the smaller, sharper answer: the gallery show in Chelsea or TriBeCa that sharpens the eye. The Morgan Library on the right day. The Frick in the right mood. The jazz club where I can slip in alone for a set and not feel out of place.

Search can give me names. AI can summarize those names.

What I want is a point of view.

For the broader seasonal view, The Culture Index maps the cultural calendar at scale. Vale is for the faster, sharper, more personal edit: the one thing worth your Saturday, the show to see before dinner, the museum to choose in the rain, the book to take on the trip.

getting dressed used to come with an editor

Fashion is no different. In some ways, it is even more complicated.

I really miss my personal shopper. Once upon a time, I had a marvelous person at Bergdorf’s who would spend half a day with me four times a year and curate all my new fashion purchases, from apparel to accessories. After my quarterly session, I would leave knowing not just what I had bought, but what outfits were meant to go where. I had a roadmap. I understood how the pieces related to one another, what the season’s logic was, and how to put the whole thing together.

I will never forget the first time my personal shopper introduced me to Alaïa. Not theoretically, but with a fit-and-flare dress that remains my go-to to this day. I wore it to my most recent book party, where the stakes could not have been higher. The fitting room itself felt like part of the experience: spacious, with flattering lighting, soft cushions on the sofa, and an ample supply of sparkling water or champagne.

It was not just shopping.

It was editing, with upholstery.

Fast forward, and I am shopping online most of the time. Meanwhile, my social and business obligations are much more complex and nuanced than they were when I was an associate at McKinsey.

It is not about looking appropriate now. It is about what I want to communicate, to whom, and at what volume level.

As a Black woman who did not grow up with parents in the corporate world, I had no one to explain any of this to me. I had to figure it out myself. The codes, the rooms, the signals, the tiny decisions that shape how you are read before you have said a word. That is still true for many high achievers without family pedigree or inherited fluency. They are expected to navigate these decisions beautifully, even when no one ever taught them how.

Those are not small distinctions.

They are the distinction.

A board lunch is not the same as a gallery dinner. A long weekend in Paris is not the same as Venice in spring. A first brunch of the season with friends is not the same as a work event where everyone insists the dress code is relaxed and no one means it.

I do not need a generic answer. I need one calibrated to the room, the city, the weather, the mood, the audience, and the version of myself I want to present that day.

That is why the question is not whether AI can show you more options. The question is whether it can refine them with judgment. I explore the style version of this problem in AI Can Pick an Outfit. But Can It Read the Room?, and the travel version in Would You Trust AI to Plan a Luxury Trip?.

the real question is always context

These seem like different problems — where to have lunch, what to see, what to wear — but they are all versions of the same one.

The question is not “what exists?”

It is “what is right for this moment, this context, this life?”

That is precisely the question most modern tools are still terrible at answering.

They are good at lists. And rankings. And confidence. They are good at serving back what is already legible on the internet. They are not especially good at tone. Or social codes. Or nuance. Also not so much at the low but meaningful hum of context that separates a decent answer from the right one.

They do not really understand that a restaurant can be excellent and still wrong. That a hotel can be beautiful and still dead on arrival. That a dress can be lovely and still say the wrong thing. That a gallery can be buzzy and still not matter.

“Best” is quite often the wrong question altogether.

The right question is: best for whom? Best for what? Best at noon on a Tuesday, or best for a first meeting, or best if the room matters, or best if you are traveling alone, or best if you want to disappear into a corner and listen to jazz without turning it into a scene?

These are not the same question, and yet our tools keep answering them as if they are.

I was not looking for more recommendations.

I was looking for a better filter.

the missing layer is taste

At a certain point, the absence of that filter began to feel absurd.

I did not need another platform to drown me in options, nor another dutiful round of “top ten” lists, another sea of tabs, another breathless stranger telling me where to go, what to buy, or why a room with terrible lighting was somehow “iconic.”

I needed the thing that used to exist, in various elegant human forms, all over my life: the trusted editor.

The friend with standards.

The personal shopper with a point of view.

The person who knew the city.

The one who had already done the sorting.

Vale began as a practical answer to an increasingly familiar problem: what happens after search has done its job and left you with too many plausible choices?

It is not meant to replace taste. It is meant to make taste easier to use at the moment you need it.

I asked Hans Hageman to build the thing I wished existed. He vibe-coded Vale, and together we shaped it into something meant to feel less like search and more like that discerning friend on call — the one with taste, context, opinions worth having, and no need to sleep.

Vale is Dandelion Chandelier’s AI taste assistant for style, travel, culture, gifts, entertaining, books, home, and the small high-stakes decisions that require judgment, not more search results.

what an ai taste assistant does

An AI taste assistant is not built to give you more options. It is built to help you choose better when taste, context, timing, and social judgment matter.

It helps answer the questions ordinary search makes harder: what to wear, where to go, what to gift, what to read, what to see, what to skip, and how to make a choice feel more considered.

Search says: here is everything.

AI says: here is everything, summarized.

Vale says: start here.

That is the difference.

Vale was built to handle the kinds of nuanced recommendations generic search handles badly. Where should I stay in Paris if I want a small boutique hotel on the Left Bank? What is the one gallery show I should see in New York this week? What is a good host gift for a dinner party at a friend’s place? Which shoe is comfortable, chic, and easy to pack for Venice? Is gourmet maple syrup worth searching for?

The point is not that Vale replaces human judgment. It is that it extends it.

It feels less like consulting a machine than texting the discerning friend who already did the sorting.

what to ask instead

The better question is rarely “what is the best?”

It is “what is right for this moment?”

Instead of asking: “What is the best restaurant in Midtown?”

Ask: “Where can I have a polished business lunch in Midtown that is quiet enough for a serious conversation?”

Instead of asking: “What should I wear?”

Ask: “What should I wear to a room where I need authority, ease, and no excess?”

Instead of asking: “Where should I stay in Paris?”

Ask: “Where should I stay in Paris if I want calm, walking, museums, and no hotel theater?”

Instead of asking: “What is a good host gift?”

Ask: “What should I bring someone who loves design, hates clutter, and already has enough wine?”

Better questions produce better answers. They also reveal the real nature of the problem.

It was never access.

It was judgment.

save this

Search gives options.

AI gives answers.

Taste gives judgment.

Before asking for “the best,” ask what the moment actually requires.

ask vale before you decide

Ask Vale one real question.

Start with the decision that is already annoying you: what to wear, where to go, what to gift, what to see, what to skip, or how to make the better edit.

Try asking:

“What is the one gallery show I should see in New York this week?”

“Where should I stay in Paris if I want a small boutique hotel on the Left Bank?”

“What is a good host gift for a dinner party at a friend’s place?”

“What should I wear to a board lunch?”

“What should I skip this weekend if I only have time for one good thing?”

Ask Vale when you do not need more options. You need a better edit.

the vale essays

Read the companion essays in this series:

AI Can Pick an Outfit. But Can It Read the Room? — what AI styling tools miss when the real question is not what looks good, but what the room requires.

Would You Trust AI to Plan a Luxury Trip? — why AI knows the five-star answers, but not always which ones fit the trip you actually need right now.

For the full product, visit Vale, Dandelion Chandelier’s AI taste assistant.

sources + further reading

• 425 Park — Midtown dining reference
  https://www.425parkrestaurant.com/
• Whitney Museum — exhibition information
  https://whitney.org/
• Nicola Vassell Gallery — contemporary art context
  https://www.nicolavassell.com/
• The Morgan Library — cultural reference
  https://www.themorgan.org/
• Bergdorf Goodman — personal shopping context
  https://www.bergdorfgoodman.com/

frequently asked questions

1. why are more options becoming less useful?

Because modern life already gives us endless options. The harder and more valuable work is choosing well: knowing what fits the moment, what deserves attention, and what should be skipped.

2. why isn’t search enough for nuanced recommendations?

Search is good at surfacing options, but nuanced recommendations depend on context, tone, subtext, and judgment. The right answer often depends on who is asking, where they are, what the occasion is, and what the answer needs to accomplish.

3. what is an ai taste assistant?

An AI taste assistant is an AI tool designed to help users choose better when taste, context, timing, and social judgment matter. Vale is Dandelion Chandelier’s AI taste assistant for style, travel, culture, gifts, entertaining, books, home, and what to skip.

4. what kinds of questions is vale built to answer?

Vale is designed for questions about what to wear, where to go, what to see, what to read, what to gift, what to bring, how to host, and what to skip.

5. can ai really help with taste-based decisions?

It can help when it behaves less like a search engine and more like an editor. The goal is not to generate more options. The goal is to narrow, sort, and return with a smarter answer.

6. what makes a recommendation feel trustworthy?

Usually some combination of context, specificity, standards, and point of view. The most useful recommendations explain not just what to choose, but why it fits the moment.

7. how should I start with vale?

Start with a real question. Tell Vale the city, occasion, mood, budget, dress code, recipient, or thing you want to avoid. The more specific the context, the better the edit.