Building in public when nobody is watching

Arman's desk in Tbilisi faced a courtyard where a man fed the neighborhood cats at 6pm every day, rain or shine. Arman knew this because for nine months he'd been at that desk from 8am to 11pm, and the cats were one of the few reliable things in his life. The other reliable thing was that his X account had 19 followers and had hovered at 19 followers since February.

He counted again that Tuesday morning, the way you check your bank account when you already know what's in it. Nineteen. One of them was his sister. One was a bot that posted about Solana. One was a guy named Devon who, as far as Arman could tell, followed every account that posted the phrase 'Day' followed by a number. So really he had 16 followers, and he wasn't sure any of them were reading.

He'd been posting like a textbook. MRR screenshots every Sunday — currently $34, one customer, a woman in Utrecht who ran a small bookkeeping practice and paid for his tiny invoicing app because she liked that it didn't try to be Xero. Week recaps on Friday. 'Day 47: shipped dark mode, talked to two users, fixed a bug in the CSV export.' Dopamine-bait threads that started with 'I made $34 in 9 months building a SaaS. Here's what I learned.' (What he'd learned, mainly, was that nobody cared what he'd learned.)

His girlfriend Nino was, by that Tuesday, tired of hearing about the account. Not unkindly. She just had the look of a person whose partner has been talking about the same thing for too long and is starting to talk about it the way you talk about a bad knee. 'Why don't you just stop posting,' she'd said the night before, not as advice, more as a question. 'If it's making you like this.'

He didn't stop posting. But he did, that Tuesday at around 10am, post something different. The dog — a small black mongrel named Beso who slept under the desk because the floor was cool — shifted against his ankle. Arman opened the compose box, deleted what he'd drafted (a week recap, of course), and wrote a real sentence instead.

The empty stadium

There's a specific kind of loneliness in building in public with a small account. You're not posting into a void exactly — the post goes somewhere, the API confirms it — but you're performing for a stadium with 16 people in it, three of whom are asleep, and you're doing the same routine the big accounts do: the confetti, the screenshots, the 'here's what I learned.' It looks identical. It just doesn't land.

The trap is that the format is contagious. You scroll the timeline, you see Pieter Levels post an MRR update, it gets 4,000 likes, and your brain quietly concludes: this is what works. So you post your MRR update too. You get one like, from Devon. You conclude: I need to post more of these. You're now in a loop where the input is 'mimic the big accounts' and the output is 'become slightly more annoying to my 16 followers, two of whom are bots.'

The thing nobody tells you is that the MRR screenshot works for Pieter because Pieter is Pieter — he's the protagonist of a story that's been running for years, his numbers are a chapter, his followers showed up for the story. Your numbers, at 19 followers, are not a chapter. They're a stranger holding up a piece of paper at a bus stop.

Arman knew this on some level. He'd just kept posting because the format was easy and the alternative — figuring out what he actually had to say — was hard.

The Tuesday post

Here's what he wrote, more or less, that morning. I'm reconstructing from memory because he later deleted it, the way people delete posts that worked, out of embarrassment that they worked.

I've got one paying customer who loves a feature I built for her — recurring invoice templates with a weird custom-field thing. Three trial users this week said the templates UI was the reason they didn't convert. Too much. They wanted one button. I keep going back and forth. Cut it and lose the customer who actually pays me, or keep it and keep losing the trials. Genuinely don't know what to do.

No screenshot. No emoji. No 'thoughts?' at the end. Just the actual problem, in the actual shape it had in his head.

He posted it and went to make coffee. When he came back there were two replies. By the end of the day there were 14, and three of them were from people with real businesses who had been through some version of the same thing. One was from a guy in Berlin who'd kept the feature and built a 'simple mode' toggle and said it was the best decision he'd made all year. One was from a woman who'd cut a similar feature and lost the customer and felt, two years later, that it was still the right call because the customer had been a tax on her roadmap. One was from someone Arman had quietly admired for a year — a founder of a small CRM, maybe 8,000 followers — who said: 'this is the most interesting post on my timeline today, please post more like this.'

Three new followers. Not 3,000. Three. But Arman looked at them and realized he knew, roughly, who they were and why they'd followed. That had never happened before.

Follower count is the wrong number

If you're running an indie account with under a thousand followers, the number on your profile is not a useful metric. It's noise. It moves on dopamine, not on substance. You'll get 40 followers from a viral take about VS Code themes and lose them all over the next three weeks because none of them actually care what you're building.

The metric that actually predicts whether your account is doing something for you is harder to name. Something like: how many people, in the last quarter, have ended up in your DMs in a way that felt like the start of a friendship? Not a 'hey loved your post' DM. A 'I'm dealing with the same thing, want to jump on a call' DM. The kind that turns into a real conversation, maybe an introduction, maybe a customer, maybe just someone who notices when you go quiet.

Call it five-DM friendships per quarter. If you've got five, your account is working, regardless of follower count. If you've got zero, you can have 10,000 followers and it still isn't.

Arman, in his nine months of MRR screenshots, had zero. The Tuesday post got him one — the founder of the CRM ended up in his DMs that evening, they talked for an hour about pricing pages, and three months later that founder would introduce him to the customer who eventually 10x'd his revenue. None of that was visible the day it happened. It just happened because he wrote a real sentence instead of a performance.

Wins versus decisions

There's a useful distinction here between two kinds of posts. The win post: 'I shipped X. I hit $Y. Here's the screenshot.' The decision post: 'I'm trying to figure out X. Here are the two sides. I don't know which is right.'

The win post is closed. It has no surface area. There's nothing to reply to except a thumbs up. People who reply look like they're congratulating you, which is fine, but it doesn't produce conversation, it produces applause. And applause from 16 people in an empty stadium sounds like rain.

The decision post is open. It has a problem in it, a real one, with stakes you can feel. People who reply are not congratulating you — they're thinking with you. They've either been through it or they're about to be. The post becomes a small room where people who care about the same problem find each other. You happen to be standing in the doorway.

Wins compound when you already have an audience. They're a reward function for people who are already paying attention. Decisions compound from zero — they're how strangers decide you're worth paying attention to in the first place.

The unwritten taxonomy

If you watch enough build-in-public accounts you can sort their posts into maybe seven or eight types. Not all are worth shipping. Here's the rough taxonomy, in descending order of how often they actually do something for a small account.

1. 1.The 'I'm stuck on X' post. A real decision in front of you, framed honestly, no fake humility. This is what worked for Arman. It's the highest-leverage post a small account can make because it pulls in exactly the people who have an opinion on X.
2. 2.The 'weird user request that's actually a feature' post. You tell the story of one specific customer asking for something strange. You explain why it's strange. You explain why, on reflection, it's not. People love these because they're little detective stories with a payoff.
3. 3.The behind-the-build post. A small, specific technical or design choice with a screenshot. Not 'here's my stack' — too generic. More like 'why I switched the empty state from an illustration to a single sentence and conversions went up.' Concrete, narrow, useful.
4. 4.The post-mortem. A real one. You shipped something, it failed, you understand why now. These are gold but they're hard — they require you to be honest about a thing you're still embarrassed about. Most post-mortems are too clean, written from the safety of three months later. The good ones still sting.
5. 5.The 'here's what I shipped, here's what I cut' post. A weekly recap but with the cuts named. Cutting things is more interesting than shipping them, because anyone can ship. Knowing what to delete is the actual job.
6. 6.The launch thread. Useful exactly once per launch. Don't pre-write it. Don't make it a 12-tweet performance. Three tweets, a link, what it does, who it's for. Stop.
7. 7.The learnings thread. 'After 6 months of building, here are 10 things I learned.' Almost always noise. The learnings are too general to be useful and too tidy to be true. Skip.
8. 8.The MRR screenshot. Earns its keep only if your number is unusual in some direction — surprisingly high, surprisingly low after a long time, surprisingly volatile — and the story around it is interesting. A boring number with a generic caption is just visual spam.

The pattern, if you squint, is that the posts that work are the ones where you're not the hero. The user is. The decision is. The bug is. You're the narrator, not the protagonist. Small accounts that try to be the protagonist read like a one-person play performed to an empty theater. Small accounts that narrate something specific and real read like a letter from a friend.

Post structures that work at zero

If you want to test this, here are five structures that consistently outperform the MRR-screenshot format when your follower count is small. None of them require you to be clever. They require you to notice something real and describe it without dressing it up.

• The honest fork. 'I'm stuck between A and B. Here's the case for A. Here's the case for B. I'm leaning toward A but I'm not sure why.' End on the not-sure. People reply because there's room to.
• The specific customer. 'A user emailed me yesterday and said X. I thought it was a complaint at first. Then I realized it was a feature request in disguise. Here's what they actually want.' One customer, one story, one shift in your understanding.
• The cut. 'I shipped this thing two weeks ago and I'm deleting it tomorrow. Here's what I built. Here's why I thought it would work. Here's why it didn't.' Self-criticism without the performance of self-criticism — just the facts of a thing you got wrong.
• The constraint reveal. 'I had two days this week to work on the app. Here's what I did with them.' Constraints make decisions visible. People relate to the constraint more than the output.
• The quiet observation. 'I noticed today that most of my paying users do this one weird workflow. I built the app assuming they'd do something else. I think I've been wrong about who this is for.' Posts where you change your mind in real time travel further than posts where you've already decided.

What changed for Arman

He didn't grow to 50,000 followers. He didn't have a viral moment. Six months after the Tuesday post he had 340 followers, which by the standards of the timeline is still tiny, and by the standards of his life was the difference between feeling like he was shouting into a courtyard and feeling like he was in a conversation.

The MRR did move — not because the posts directly converted, but because the posts pulled him into a small loose network of other indie founders who talked to each other in DMs, who introduced him to customers, who told him when his pricing page was bad and when his copy was confusing. The CRM founder from Berlin became a friend. The woman who'd cut her own feature became, eventually, his second-biggest customer (she ran a small agency that needed exactly what he was building). The 14 replies on that Tuesday were the first thread in something that turned into a tiny scene.

He kept the templates feature, by the way. Built a 'simple mode' toggle, mostly because the Berlin guy had told him about doing the same thing, and the trial conversion rate climbed from 4% to 11% over the next two months. The Utrecht bookkeeper sent him a thank-you note when she discovered toggling out of simple mode brought her templates back. He printed it out and stuck it on the wall next to his desk, above where Beso slept.

What changed wasn't really the strategy. It was the unit of work. He stopped trying to write posts that would 'do well.' He started trying to write posts that he'd want to read if someone else posted them. Almost every post got worse on the surface — less polished, less framed, sometimes just a single sentence about something that bothered him — and almost every post did better on every metric he eventually decided to care about.

Who to read when you're starting

The other thing that changed: Arman started paying attention to a different kind of account. Not the ones with the biggest numbers. The ones whose posts he actually wanted to read on a Wednesday morning when he was tired. He made a private list. It had maybe 30 people on it after a few months — small founders, designers who wrote about their work without hyping it, two or three writers, a former Stripe engineer who posted exclusively about hiring, an illustrator in Mexico City who occasionally wrote about pricing.

None of them were 'build in public' influencers. Most of them had between 500 and 8,000 followers. All of them posted the way Arman was learning to post — specific, undefended, in their own voice. When he later helped a friend start an indie account, the first thing he did was send her a version of that list. Eventually the list ended up as the seed of the follows directory on onedb, which is now where Arman tells people to start when they ask him who to read.

It's not a list of who to copy. It's a list of who to be near. Those are different things. You copy a voice and your account sounds like a cover band. You stand near a voice and your own voice gets a little braver.

The thesis

When nobody is watching, the only useful post is one you'd want to read yourself. That's the entire thesis. Everything else — the cadence, the formats, the metrics, the threads, the optimal time to tweet — is downstream of that single test. The Tuesday post passed it. The 200 posts before it didn't.

Building in public, at the small-account stage, isn't a distribution strategy. It's a way of thinking in public, and the people who'll matter to you over the next two years are the ones who recognize their own thinking in yours. They aren't going to find you through an MRR screenshot. They're going to find you through a sentence that sounded like it came from someone who actually had the problem they have.

If you've been posting Day-47 updates for nine months and you're stuck at 19 followers, the answer isn't more updates. The answer is one honest sentence about the thing you actually thought about in the shower this morning. Post that. See who shows up.

The man in Arman's courtyard still feeds the cats at 6pm. Beso still sleeps under the desk. Nino doesn't ask him to stop posting anymore — partly because he posts less, partly because when he does post she sometimes reads it over his shoulder and says, 'yeah, that one's good,' which is more than he ever got from a screenshot.

The follower count is still small, by internet standards. But the stadium isn't empty anymore. There are maybe 30 people in it now, and he knows most of their names, and on a good week one of them sends him a DM that starts with 'I've been thinking about what you said' — and that, it turns out, was the whole thing he was trying to build all along.