Finding ideal customers for a daily build-in-public tool

Everyone tells you to find your ideal customers. Nobody tells you where they are at four o'clock on a Tuesday. The honest answer is that they are typing the problem into a social feed, in their own words, with no idea a product exists for it. Someone writes that they spent hours editing a video and then never posted it. Someone else writes that their one regret is not documenting their wins as they went. Those are buying signals, and they scroll past you because you were building.

The second half of the problem is scale. Searching five platforms by hand, every day, with the right phrasing on each one, is a job. Do it manually and you either skip it entirely or you burn an hour reading posts that turn out to be irrelevant. The thing that kills it is the same thing that kills documenting: friction.

That is why I built this into the same app I document in. For me, this whole app is you're documenting on one side, and you've got your promotion and marketing channels on the other side, and it makes both incredibly simple for you. One place to record what happened today, one place to go and find the people who need what you are making.

In Underway, your episodes are the diary. Promote is the shop window: the things you make to be found and to be run. Nothing in there is numbered, nothing counts towards your streak, and none of it reaches your interviews. It is deliberately kept apart from the record you are building.

The shop window holds your marketing brief, written from your site and your episodes and yours to correct. Found in AI answers, which every week puts the questions your buyers ask to ChatGPT, Claude and Gemini and tells you whether the answer names you or a competitor. Videos to film, what people search for on YouTube. Library, every video you have filmed and what each one is doing. Ad answers, marketing assets, and Reach, which shows what the posts you sent did on each network.

Conversations is the one marked part of series, on your own key. Found in AI answers and Conversations are both on the top membership. Everything else in the shop window sits alongside them, so the day's reply list and the day's episode live in the same app.

Every day, on your own key, it searches X, LinkedIn, Reddit, Threads and Hacker News for people saying, in their own words, that they have the problem you solve, and sorts them by platform. Each one comes with the link to the post, the context, how it was found and what you could say. You open it and start the conversation. Nothing is posted for you, and your interviews never see it.

Open the page and it fills as each platform finishes, so you can leave it and come back. On the run I filmed, X finished first with 28 to read, then LinkedIn done with 13, then Reddit with 33, with Threads and Hacker News still waiting. A progress bar moves through them. One of the platforms came back with Couldn't search, which happens.

The search itself is built platform by platform: one moment, said a few ways, each in quotes and joined together. On X that was people saying they wish they had filmed, wish they had recorded, wish they had documented, should have filmed, or that editing takes forever, or that they never know what to say. On LinkedIn, wish I had documented and struggling to post consistently. On Reddit, building in public. It is not always spot on, but people saying that kind of thing are probably close to what I do.

It does this by itself, which is exactly why the setup matters. Do the marketing brief, give it your website, give it as much context as possible so it understands what your buyers' issues are and what to look for. The search terms are written from that brief. Vague brief, vague search.

Scanning five platforms produces far too much. The second pass is the one that matters: it reads what it found and decides whether the post is actually worth replying to, because the context has to be there, not just the phrase. On X it told me plainly that it looked at 13 more and passed over them. On LinkedIn: looked at 18, read 13, 6 worth a reply. Replies were written for 25 of 43 posts on that run.

You still have to use your judgement. The first thing it surfaced that day was worth a reply but was clearly a competitor: someone who hates editing video so they built their own tool for it, 39 seconds in and 18.6 seconds out. Look, this tool is not going to be spot on every time. Some of the people it finds will not be founders at all, which does not mean Underway cannot help them.

Each signal is labelled with the kind of moment it is: frustrated with it, doing it the hard way. One from X read that they spent hours editing a video just to not post it, tagged as the edit-kills-it moment worth a practical reply, found by searching X for a cluster of editing complaints. Another said sorry this is 16 minutes long, not offended if no one watches, I just hate editing. These are not leads in a CRM, they are people mid-complaint.

Under each post it drafts what you could say, with a character count against the platform's limit, and the rule printed underneath: helpful, no link, no product. For the person who spent hours editing and never posted, the draft said the edit is usually where it dies, that posting the raw take the same day tends to help because the version you edit for hours is the one you judge too hard to ship, and ended with a question: how long had it been sitting.

You get three buttons: Reply on X, I replied, or Skip. Press reply and it carries the words through to the platform, where you post it yourself. It does not post on your behalf, and none of this reaches the interviews that make your episodes.

The point of a reply with no link is that it has to survive on its own. You are adding something practical to a thread, not pitching. If the person wants to know who you are they can look, and that is a far better first contact than a cold message.

The one that made the run worth filming came from LinkedIn. Alan had written about 2018, $326 in the bank, broke, rejected, ghosted after seven interviews, final rounds and positive feedback and then we'll get back to you. He had closed deals, hit quota, built pipeline, but no big logos on his resume. Just results, and apparently that wasn't enough.

The context Underway put under it said his one regret is not documenting his wins as he went, which is exactly the record-as-you-go problem, and that there is something practical to add about capturing it in the moment rather than reconstructing later. Found by searching LinkedIn for wish I had documented. That is literally my exact customer.

The drafted reply said the hard part with documenting is that the moment you'd want to capture is always the moment you're busiest, that a fixed two minutes at the end of the week saying out loud what happened tends to work, and that reconstructing a year later is where most people lose it. Then a question: did you ever manage to rebuild any of those early numbers.

On Reddit it found someone nine months into building a trading alert bot, posting a long written progress update as a follow-up to an earlier one, flagged as the slow manual way of keeping a record of building. Same problem, different words, different platform.

The search runs itself while you do something else. Your time goes on reading the shortlist and sending the replies you agree with, which is minutes rather than an afternoon. Yes, you are paying for Underway, but this is semi-free marketing. Doing it every day, across all of these signals from people saying they have the problem you have a solution for, is how finding your first few customers can be done.

Compare that with the usual options. Cold outreach to people who never said they had the problem converts badly and makes you feel awful. Paid ads work but cost money every day and you are guessing at the audience. Posting into your own feed and hoping reaches whoever already follows you, which when you are starting is nobody. Hiring someone to do the searching gives you a second person's guess at your buyers.

It suits anyone selling to people who talk in public about their work: founders, makers, anyone whose buyers post on X, LinkedIn, Reddit, Threads or Hacker News. It suits you less if your buyers never post, or if you want a tool that fires off messages automatically, because nothing here is sent for you.

I work in marketing and sales, so this part comes naturally to me. What changed is that the documenting now happens in the same place and takes no more effort. Underway reads your day from the tools you already use, interviews you out loud about what changed, and hands back a finished numbered episode with no editing. Then the Promote tab takes the same understanding of your business and goes looking for people with the problem.

The two halves feed each other. The marketing brief that gets written from your website and your episodes is what the search terms come from. The episodes you record become the thing a person can go and watch after you have replied helpfully to their post. Neither works as well alone.

This is a one-stop shop to help you document your business and also propel it forward in terms of promotion, marketing and finding customers. That is what finding ideal customers looks like for me at this stage: a daily shortlist of real people, in their own words, with something useful to say back.