Every time I answer this question, I feel I can only provide part of the answer. My thoughts are fragmented. So I decided to document them here and clearly show the difference.
Several months ago, I joined a WhatsApp channel with around 500 people. When a community becomes big enough, you start noticing interesting patterns.
One thing I noticed is this: no matter who announces that they are going to launch a product in the channel, everyone usually follows the same pattern. “That sounds amazing.” “I would definitely use it.” “Can’t wait to try.”
Of course, I do not think people are intentionally lying. Most of the time, people are just being kind. They do not want to be the person who sounds negative.
But I also think many people do not even notice this coordinated behaviour at all. This is just the most natural reaction towards another person’s idea.
And I think this is not only a community problem. This is a human reaction. It is much easier to agree than to challenge.
I saw a very hot post on Reddit asking: Why do people suddenly feel they have so many competitors the moment they start doing marketing?
I think this question is very interesting. When you are still building, the world feels quiet. You are focused on your own product, your features, your roadmap. You may even feel your idea is quite unique.
But the moment you start marketing, life suddenly becomes harsher. Because now you are not only building anymore. You are trying to sell.
And when you try to sell, you are forced to look at the market for real. Suddenly, competitors pop up everywhere. Fear starts to creep over you.
And you start wondering: why do all these competitors appear exactly when I finally start marketing? But maybe they were always there. Your brain just strategically avoided seeing them before. Marketing removes the illusion that building alone is enough.
And this is also why using a general LLM directly can sometimes become tricky.
In artificial intelligence, there is a concept called AI sycophancy. It means that large language models sometimes tailor their responses to what they think the user wants to hear, instead of what is accurate, useful, or warranted.
The behaviour can take many forms. An assistant may agree even when your opinion is too weak. It may abandon a correct answer after you ask, “Are you sure?” It may validate your belief, your decision, your product idea, or even your self-image too quickly. It may praise your work in a way that feels good, but does not actually help you see the truth.
Does this behaviour sound similar to what I described earlier in the WhatsApp channel? I think it does. In both cases, it is a very human reaction.
I published a post a week ago about how to quickly find your first digital product to sell on Gumroad using the famous product design framework, CIRCLES.(https://xianli.substack.com/p/how-to-use-the-circles-framework)
The feedback was huge. People kept telling me it was useful.
You can probably get similar results after chatting with an LLM for hours. But the key word is hours. Using a framework early can accelerate the whole development process.
As a developer, you can definitely build a project from scratch by writing the code line by line. However, most of the time, we still choose a framework because it helps us build faster and more consistently.
The same applies when building with the API.
I can treat the model as part of a structured product, not just a friendly chatbot. I can give it stricter rules. I can force it to judge your idea through a clear framework. I can ask it to surface red flags, compare your idea against failed products, and score it with clear criteria.
In the early stage, encouragement is easy to find. But clear judgment is much harder to find.