Why We're Rethinking Product Discovery

Search bars have been the front door to online shopping for over two decades. Type what you want, scroll through results, hope for the best. It worked when there were a few hundred products to choose from. It breaks down completely when there are millions.

The problem isn't that search engines are bad at finding products. They're remarkably good at matching keywords to listings. The problem is that shoppers don't think in keywords. You don't wake up and think "men's lightweight moisture-wicking running shirt blue medium." You think "I need something to run in this summer that won't feel disgusting after three miles."

That gap — between how people think about what they want and how search engines expect them to ask for it — is where most of the frustration in online shopping lives. Every abandoned search, every "I'll just go to the store," every impulse buy that ends up in a returns box can be traced back to that same disconnect.

The filter problem

The industry's answer has been filters. Price range, brand, color, size, rating, material, shipping speed — stack enough filters and eventually you'll narrow down the results. But filters assume you already know what you're filtering for. They work for replacement purchases ("I need the exact same running shoe in a new size") and fail for discovery ("I'm getting into trail running and have no idea what I need").

Filters also create a paradox: the more options you add, the more work the shopper has to do. Every dropdown menu is a decision. Every decision costs mental energy. By the time you've configured seven filters, you've spent more effort searching than you would have spent walking into a store and asking someone for help.

What asking someone for help actually looks like

That store analogy is important. When you walk into a good running store and say "I'm getting into trail running," the person behind the counter doesn't hand you a spreadsheet. They ask questions. What terrain? How far are you running? Do you have any knee issues? What's your budget? They use your answers to narrow down options, explain the trade-offs, and make a recommendation.

That interaction is everything online shopping isn't: contextual, conversational, and adaptive. It starts with your intent and works backward to the product, instead of starting with the product catalog and hoping you can find what matches your intent.

This is what we're building

Starseek started with a simple question: what if online shopping worked like talking to that person at the running store — except they had access to products from every store, not just their own?

Our AI assistant, Spark, is built around this idea. You describe what you're looking for in natural language — as vaguely or specifically as you want — and it has a conversation with you. It asks clarifying questions. It understands context ("something for summer" means lightweight and breathable). It searches across 11 e-commerce platforms simultaneously and brings back results that match what you actually meant, not just what you typed.

But the real unlock isn't the AI conversation itself. It's what happens over time. Every interaction teaches the system something about your preferences — your price sensitivity, your brand affinities, your style, your values. The third time you use Starseek, it's noticeably better than the first. By the tenth, it starts surfacing things you didn't even know you wanted but are exactly right.

Discovery, not just search

There's an important distinction between search and discovery. Search is "I know what I want, help me find it." Discovery is "I don't know what I want yet, help me figure it out." Most e-commerce platforms are built for search. Almost none are built for discovery.

Discovery requires understanding the person, not just the query. It requires connecting products across brands and platforms. It requires an AI that can hold context across a conversation and learn from patterns across many conversations. It requires a fundamentally different architecture than the keyword-matching systems that power online shopping today.

That's the bet we're making. Not that we can build a better search bar — but that the search bar itself is the wrong interface for how people actually want to shop.

We're not there yet. We're still building. But every week, the gap between "good store employee who knows you" and "AI that learns your taste across every brand on the internet" gets a little smaller. And that's what gets us out of bed in the morning.