Why an answer isn’t always the answer
When I tried Googling “what is an answer,” I got: “Please provide the specific question, puzzle, or problem you are trying to solve so I can give you the right answer.”
It seems answers are defined only in relation to their questions. There is no definition of an “answer” on its own.
I looked into this and found plenty of philosophy on “what is a question?” and far less on “what is an answer?” I came across mathematics papers on how to get to a solution, which is related, but still focused on the question. Searching for an answer on “answers” is a lot harder than it sounds.
WHEN CONTEXT DILUTES THE ANSWER
This becomes interesting in a world where answers are now delivered at scale. We’re getting extraordinarily good at producing them, but a gap remains. We don’t really know what an answer is. Basing an answer on the question is no longer enough.
AI has dramatically increased our ability to find answers quickly and provide a wider aperture of information. Competition for this service has shifted into which tools can provide the best, fastest, and most accurate answers.
We may be able to determine “fastest” and “most accurate,” but how do we determine “best?” Is it correct? Relevant? Complete? That works when a question has a definitive endpoint or when I am looking for a fact. I can ask, “How far is the moon from Earth?” get a number, and move on. Chatbots do this well. Often, the one-sentence answer is perfect, and adding too much context around the distance to the moon dilutes the information I need.
Other times, the answer is the first thing I need, not the endpoint. “How far is the moon from Earth?” can lead to “How long does it take to get there?” Now the answer depends on how, who, and why. It gets more complex. It gives me enough context to ask the next question, compare sources, explore something I didn’t understand, form a point of view, or build something from it.
If every answer is optimized to close the tab, what happens to the answers that are supposed to start something?
A BUILDING PROCESS
Building knowledge is a process. Sometimes it’s serendipitous discovery; other times, it’s methodical.
A good answer depends on what you are trying to do with it. Correctness is necessary, but not always enough. It needs context, comparison, sources, or follow-up questions.
So when is an answer best? Is it a human source? Is it longer? Does it have more citations? There still is no clear answer here. It depends on what I need.
I view this as the pipeline: Information → answer → understanding → application.
The answer is not the end goal; it’s merely the midpoint. Answers unlock deeper understanding and, ultimately, application. AI has made the first half of this pipeline wildly easier. That’s both a gift and a risk: it can dull the curiosity to dig deeper. Beyond a quick answer is how other people have approached the question before, what they asked next, and what they did with it.
I find the back half much more interesting. But it’s more difficult, too. Understanding is subjective. Application is contextual. Both are much harder to measure than whether a system returned the correct answer.
I think a lot about this at Scribd, Inc. We are a knowledge company, and we want to help people get to the second half of that pipeline. But that does not mean we skip over the answer. You can’t get to application without it.
I don’t know if anyone can “win” this. I don’t even think anyone can quite define it yet. Google was right in a way. It couldn’t tell me what an answer was without knowing what I wanted it for.
The “best” answer is the one that meets your objective, whether that’s finding a fact or learning something new. We’ve gotten very good at the answers that provide a fact. We are just scratching the surface on the ones that open something new.The next step is “beyond the answer”—what it allows someone to learn, create, or do.
Tony Grimminck is CEO of Scribd, Inc.













