Loading page…
An API that turns any URL into a screenshot — one boring feature, 280 paying companies, about $12K a month
280 customers · plans from $17/mo · 40–60% profit margins · ~$4,500/mo total costs · ~2M screenshots rendered per month
He deliberately shrank his idea list until only one shape was left: "I reduced my list of ideas only to API products," because building APIs was the one thing he already did better than most people.
Fewer bars = easier, cheaper, or faster for an AI-assisted solo builder. Editorial judgments based on the case details.
Dmytro spent most of his working life as a backend developer inside startups and companies, building products for other people. By his own account he had a good job and a large salary, and still felt pushed in another direction — becoming an independent builder was a childhood ambition. Two years before this interview he quit, and started experimenting. The first thing he did was not brainstorm harder, it was cut. He says he had millions of ideas, and understood that the real constraint was his own ability to ship something of quality alone. So he narrowed the space deliberately: "since I was a backend developer, server side developer, I knew how to build APIs, it was kind of my super skill — I reduced my list of ideas only to API products." Two candidates survived. One was an email-validation API, which he actually built and then found boring. The other was a screenshot API. He checked Google, found a crowd of existing screenshot tools, and treated that as a buy signal rather than a warning: competitors mean people are paying. The first version took five months because he wanted it perfect — a mistake he now names openly. Launch produced exactly one customer at $7 a month, which he describes as feeling like a million dollars. Two years after quitting he was at 280 customers and roughly $12,000 a month in recurring revenue, on plans starting around $17 a month with usage-based upsells for extra screenshots and features.
The first version took five months, which he now calls a mistake — "I wanted to be everything perfect," and "your first version will be anyway not good." Then came a month of manual promotion across every platform he could reach: Twitter, Reddit, Indie Hackers.
The quick-reference card — acquisition channels, replication playbook, risks & traps, and our verdict — is members-only.
Join the Unicorn Club — $5/moCancel anytime, in one clickPaste into Claude Code / Codex and get a working version of this product end-to-end.
Two production-grade prompts per idea — a full build spec and an SEO growth plan — plus 10 tailored build specs a week. Members only.
Join the Unicorn Club — $5/moCancel anytime, in one clickEvery claim traces back to a listed source — open any of them and check the number yourself. The exact words are quoted below, with a timestamp that jumps to the second each figure is said.
“…screenshots So currently I am at 280 customers making around $12,000 mon recovering Revenue this is subscription model you subscribe to a product and pay for example $17 a…”
1:51 · hear it in the video“…4,000 per month for servers total expenses should be around $4,500 I think like in total that's awesome so we already kind of talked about how having competitors is…”
10:58 · hear it in the videoData credibility: Founder interview on Starter Story (host Pat Walls). Every number — 280 customers, ~$12,000/mo recurring, plans from $17/mo, 40–60% profit margins, $3–4K/mo server cost, ~$4,500/mo total expenses, ~2M screenshots per month, churn from ~11% down to ~7% — is stated by the founder in conversation. No dashboard, invoice or third-party data appears in the transcript, so treat all of it as self-reported and unaudited. The founder's first name is rendered by the auto-caption as 'demitro' / 'mro'; the product is ScreenshotOne and the founder goes by Dmytro, but verify the exact spelling before publication. No country or city is mentioned anywhere in the transcript, so we have left region unset rather than guessing from the name. The tech stack (TypeScript + Puppeteer, Go, Cloudflare, PostHog, Crisp, Google Search Console) and the marketing channels are as described by him.
One person, $7 a month, after thirty days of posting. "For me it was a huge feeling, it felt like a million dollars."
Three loops started compounding, and none of them needed him to keep posting. Google is the main one, run deliberately rather than hopefully: Search Console for keywords, click-through and positions, Keyword Planner to size a term's volume before writing anything for it.
The silent stretch was six months: five spent building a first version he wanted to be perfect, plus the month of posting everywhere that produced a single $7 customer. Two things told him to continue rather than quit.
ScreenshotOne is a hosted API that automates screenshots. A company sends a URL and parameters, and gets back an image. That is the whole product surface. The customers are companies from small to large that need screenshots generated programmatically — think link previews, social cards, PDF and thumbnail generation, monitoring, and automated reporting. Critically, he positions it as a developer product and says so loudly. He found that some signups arrived expecting a no-code tool, churned when they discovered it wasn't, so he rewrote the copy to emphasise that you need to write code to integrate it. That is a positioning fix, not a product fix, and it is one of the more useful details in the interview. The hidden depth is operational. Behind the simple endpoint sits a fleet of headless browsers rendering roughly two million screenshots a month. Our read: this is the honest shape of "simple tool" businesses at the API layer — the interface is trivial, the reliability is everything, and the difficulty lives in queueing, rate limiting, browser crashes, fonts, cookie banners, lazy-loaded images and per-customer isolation. A weekend clone gets you the demo. Getting to two million renders a month with customers who trust you is the actual work, and it is why the moat is real even though the feature is not novel.
Straight subscription with usage expansion: you subscribe, pay something like $17 a month, and pay more if you need extra screenshots or extra features. Two hundred and eighty customers produce about $12,000 a month, which averages out to roughly $43 per customer — so the account mix clearly skews well above the entry tier. The pricing history is the useful part. He launched at $5, felt that was too cheap, went to $7, and discovered his margins were "almost zero" — a screenshot API has genuine per-request cost, so an underpriced plan is a loss-making plan. His fix was to shrink the free plan and raise prices on the paid tiers. His observation about what happened next is worth stealing: "when you raise prices it really helps, it's a much stronger signal to get people who really want to use your product." He describes current pricing as intuition — how much customers can still pay while the value holds and the margin works. Margins run 40 to 60 percent. Servers are the dominant cost at $3,000 to $4,000 a month, with total expenses around $4,500. Our read: that cost structure is the single biggest difference between this and a pure-software SaaS. You cannot price this like a $9 productivity app, because every request burns CPU and bandwidth. Anyone copying an infrastructure-shaped API needs the pricing model wired to usage from day one.
He tried nearly everything at the start and settled on two majors: Twitter and Google. Google is the compounding one — he uses Search Console to track keywords, CTR and positions, and Keyword Planner to size volume before writing content. For a technical product with clear job-to-be-done queries, that is a durable channel. The more interesting answers are the ones he calls "unobvious channels." Zapier and Make are the standout: enormous numbers of businesses automate work inside those platforms, and publishing your product as an app inside them means people find you while shopping for a step in their workflow, not while searching for a vendor. Product Hunt gave him an awareness spike and, he says, an SEO boost from the backlink. The accidental channel is the best story. Someone published a YouTube tutorial on sending cold emails using his product, and he watched paying customers arrive from it. His conclusion was to make YouTube a deliberate channel: publish tutorials for a technical product, and get double exposure because Google surfaces video results for keywords where ranking a page on position one is brutally hard. He calls it a hack; it is really just recognising that YouTube is a search engine with less competition for developer queries. The first dollar came from a month of manual promotion everywhere — Twitter, Reddit, Indie Hackers, every platform he could reach — which produced exactly one $7 customer.
The stack is fully disclosed and unglamorous. TypeScript manages the headless browsers via Puppeteer. Go handles rate limiting and API key management — a deliberate split, with the performance-sensitive gatekeeping layer in a different language from the orchestration layer. Cloudflare stores the screenshots and fronts the API. On the business side: Google Search Console for keyword, CTR and position analysis, Google Keyword Planner for volume estimates before writing content, PostHog for product analytics and funnels — specifically so that when a payment lands he can see which marketing channel produced it — and Crisp for live chat support, wired to his phone so he can answer and often resolve issues immediately. Costs: $3,000 to $4,000 a month on servers to render around two million screenshots, roughly $4,500 a month in total expenses, and 40 to 60 percent profit margins on about $12,000 in monthly revenue. Our read: the margin band is the number to internalise. Most indie SaaS advice assumes 90 percent-plus gross margins, and that assumption breaks completely for compute-heavy products. At 40 to 60 percent, a pricing mistake or an abusive customer is not an annoyance, it is a hole in the business. The upside is that the cost floor is also a competitive moat — a hobbyist can clone the endpoint but will discover that running browsers reliably at volume is a full operational discipline, and that's before you handle the customer who points your renderer at ten thousand URLs an hour.
Copy the idea-selection method first, because it is the most transferable thing here. He did not look for the best market; he looked for the intersection of what he could uniquely build well and what already had paying demand. "Reduce your space of ideas to something that I can really build" — for him, APIs. Then he validated the niche by looking at competitors and their customer counts, on the logic that a crowded search page proves budget exists. His validation bar is worth writing down: not opinions, not pre-orders from friends, but ten paying customers from outside your network who didn't know you and use it for their business. Copy the churn work too. When customers cancelled, the cancellation-reason dropdown gave him nothing useful. So he researched each churned account — their profile, their company, their likely use case — then emailed a single yes/no guess: did you cancel because of pricing? He says he got close to a 100 percent response rate with that approach. It is pure manual labour and it built the mental model that took churn from about 11 percent monthly to close to 7, with a target under 5. What to change: don't take five months to launch. He now aims for a month or less, shipping one feature with no auth and no payments, having friends click through it, then launching on Twitter and reaching out to potential customers directly. And if you copy the compute-heavy API shape, price for usage from the first day rather than discovering zero margin at $7.
Server cost is the structural risk. At $3,000 to $4,000 a month against $12,000 in revenue, a hosting price rise, an inefficient render path, or a single customer hammering the API compresses margin immediately. Compute-heavy businesses have no cushion. Competition is heavy by the founder's own description — he found many screenshot APIs on Google before he started, and the barrier to shipping a basic version keeps dropping. His answer is to serve the customers he has extremely well rather than fight on features, which works at $12K a month but does not scale into a defensible position if a well-funded entrant decides to give screenshots away as a loss leader. Platform dependency runs deep and mostly invisibly: the product is headless Chrome. Browser changes, anti-bot defences, Cloudflare-style challenges on target sites, and site owners who don't want to be rendered all sit outside his control. A meaningful chunk of support load in this category is "your API can't screenshot this specific page." Churn was 11 percent a month and is around 7 — better, but that still means the customer base turns over fast enough that acquisition can never stop. Finally, it is one person. He is the developer, the marketer, the support desk on his phone, and the SEO team. He talks sensibly about protecting his mental health precisely because the business has no redundancy. And the evidence is his own: 280 customers, ~$12,000/mo, 40–60 percent margins and the cost figures are all self-reported in an interview, with no dashboard shown in the transcript.
The four founder playbooks — first customers, first dollar, the flywheel, the silent stretch — are in the Unicorn Club, on this idea and on all 1,435 others. The story, the numbers, the sources and every later section stay open to everyone.
Join the Unicorn Club — $5/moCancel anytime, in one click