Talk to users, and how to find them with no network

Because the advice skips the three hard parts: which users, found where, asked what. Building has a clear to-do list and conversations do not, so many founders build another feature instead.

Vibe coders are people who build software mostly by prompting artificial intelligence (AI) tools. Many of them say some version of "I can build, but I can't sell." That is our reading of founder communities, not a measured ranking. Our view: behind that feeling there are usually very few real customer conversations.

Not "anyone who would use this." A specific person with a specific moment of pain you can name.

Here is an illustrative example, not a real founder or a real tool run. Say you are building a tool that helps physical therapists produce superbills. A superbill is a detailed receipt a patient sends to their insurer to get money back. "Physical therapists" is far too broad.

A sharper target: the owner of a small cash-pay practice, where patients pay directly rather than through insurance. They are one to five years in, and they handled a patient's paperwork request themselves in the last month.

Now name the wrong people out loud. Therapists who bill insurance directly do not have this problem. Owners who handed the paperwork to an assistant long ago no longer feel it.

That cut changes everything that follows: the message you send, the questions, what a warning sign sounds like. Here is a simple test of your customer definition: is the wrong person easy to name? If nobody is clearly outside it, you do not have one yet. If you struggle to describe the person at all, sharpen the problem into one sentence first.

In the places they already complain. Most trades and hobbies have forums, online groups, professional communities and comment threads. Read a few weeks of recent posts before you message anyone. You will learn their words for the problem, and whether it comes up at all.

Reach out to people who described the problem in public. Mention the specific thing they wrote. Ask for twenty minutes about how their week actually works, and promise no demo or sales pitch. People are more willing to answer someone who clearly read what they wrote. That is our judgment, not a measured reply rate.

Here is an illustrative message for the therapist example. Swap in your own details:

Hi Sam, I saw your post about chasing superbill requests between sessions. I'm researching how small cash-pay practices handle that paperwork. Could I ask you about how your week works, for twenty minutes? I'm not selling anything, and I'll send you a short summary of what I hear from other owners.

Keep it short, specific and easy to say no to. Offering something back, like that summary, gives people a reason to say yes.

The Customer Discovery Accelerator helps draft this plan for your idea: whom to target, whom to leave out, where they may gather, a first message and questions to ask. Treat its suggestions as a starting point and check them against what you find in those communities.

Ask about past behavior, never future intention. "Would you use this?" invites a kind yes that costs the person nothing, a point Rob Fitzpatrick's book The Mom Test (2013) makes at length. Ask instead: walk me through the last time this happened, what did you do next, what have you tried, what does it cost you now?

In the therapist example, a good opening question is "Walk me through the last time a patient asked you for a superbill." It never describes the tool. Your idea comes up only near the end, when you ask whom else you should talk to. A real name offered on the spot is a signal. A polite "sounds useful" is not.

For a full question set and how to fix leading questions, read what to ask in a customer interview.

It can be. That is the strongest case against this advice. Some products are best tested with a small working version, because people cannot judge them from a description. Interviews also do not replace shipping. A working product with real use tells you more than any conversation.

The missing fact is commitment. Has anyone offered time, money or an introduction? Once a few people commit something, move to the smallest thing you can build to test it.

Keep going until you hear the same pain, told the same way, from people who do not know each other. One enthusiastic talk is an anecdote. A frustration you hear again and again, in more than one community, is a direction.

Write two sentences: who your person is, and who is left out. Find one community where that person posts. Read recent threads, then message three people who described the problem, each mentioning what they actually wrote.