I built a SaaS in 10 days. 3 weeks live. 0 paying customers. Here's the data.
PostDew is a LinkedIn post generator that strips the AI rhythm out of generated content. The "let's dive in", the arrow bullets, the parallel structures, the ChatGPT cadence everyone can smell now. I built the MVP in 10 days around my full-time dev job, deployed to a 5 euro Hetzner box, shipped 21 days ago.
The numbers right now:
- 0 paying customers
- ~10 cold DMs sent on LinkedIn (free tier limited me)
- 2 replies (one polite decline, one polite thanks)
- 0 signups from those replies
- Submitted to HN yesterday, got flagged, 30-40 visitors, 0 signups
- I use PostDew myself daily to post on the company LinkedIn page
What I'm learning the hard way: building and selling are completely different sports. I treated the build like a sprint and marketing like an afterthought. That's the inversion most indie devs make.
The meta-irony I keep noticing: I built this to help people who don't want to write LinkedIn posts. Now I'm writing LinkedIn posts about not having customers for the tool that helps you write LinkedIn posts. The dog-fooding is real. The validation that anyone wants this besides me is not.
What I've tried:
- Cold DMs to 10 LinkedIn operators. 2 replies, 0 signups.
- One Show HN. Flagged within hours.
- One vulnerable LinkedIn post about all of this. Crickets.
What I'm doing this week:
- Burning a LinkedIn Premium free trial on 50 DMs to people who self-identified the pain in a recent post ("haven't been posting", "trying to post more")
- Cross-posting this story here and on Reddit
- Recording a YouTube vlog about it
What I'd value feedback on:
- Anyone shipped a content tool in this category and survived to find your first 10 paying users? What was the unlock?
- Is "0 customers in 3 weeks" panic-mode or statistical noise? I keep telling myself noise but I've never been here before.
- What's the AI tell that makes YOUR skin crawl on LinkedIn that I should add to the strip-list?
Tech stack: Django + DRF + Celery + Redis on Hetzner CX23, React 19 + Tailwind v4 on Vercel, Supabase Postgres, OpenAI gpt-4.1-mini default with Anthropic fallback, DodoPayments for billing. Whole infra runs on roughly $5/month.
Demo (no signup, 1 free per IP per week): postdew.com
Brutal feedback welcome.
About to launch a finance app myself, so the HN flag detail caught my eye. Did you ever figure out what triggered it (thin profile, link-in-title pattern, account age)? The Show HN guidance feels like a moving target right now and I'd rather not learn it the hard way. The infra split also lands for me: I'm on Vercel, DynamoDB and Stripe targeting under £5/month, and I've come round to the view that the lean stack stops being a flex once you realise distribution is the actual bottleneck, not hosting cost.
3 weeks, 0 paying customers — that's noise, not signal. The clock starts when you have real distribution, not deploy day.
On the AI tell question, the ones that make me close the tab immediately: the "one sentence. per line. for. dramatic. effect." formatting, any post that opens with "I almost quit.", "Here's what nobody tells you:" (everyone tells you this), and the rhetorical question as hook / answer as punchline / CTA as closer — same structure, every post, every account.
The rhythm problem is actually structural, not lexical. The words can be fine but the sentence-length pattern gives it away every time.
On distribution: cold DMs to strangers who haven't felt the pain recently won't convert. Your best channel is probably the comment section of posts where someone is actively complaining about sounding like AI. Inbound intent beats cold push every time.
this is the most useful comment i've gotten on any of these posts.
three things i'm doing this week off your feedback:
also taking the "noise vs signal" reframe seriously - i've been measuring against deploy day, not distribution day. fair point.
if you ever want a free pro account to keep poking holes, let me know. genuinely useful read.
The product problem is real.
But I honestly think “PostDew” is quietly hurting you before the output even gets judged.
You’re building for operators and founders who are already hyper-sensitive to AI slop, generic tone, and low-trust tooling.
Then they land on a name that sounds lightweight, temporary, and very close to another “AI post helper.”
That mismatch matters more in this category than most people think.
Especially because your actual angle is stronger than “LinkedIn post generator.”
You’re really selling:
→ de-AI-fication
→ natural rhythm restoration
→ believable writing
→ protecting credibility online
That’s a much more serious positioning layer than the current name communicates.
Honestly, something like Beryxa.com, Xevoa.com, or Exirra.com would frame the product closer to an operator-grade writing system instead of another content utility.
The painful part is:
you may actually be solving a real problem already — but the brand currently feels easier to ignore than the problem itself.
appreciate the read.
you're right that the angle (protecting credibility, natural rhythm restoration) is stronger than "LinkedIn post generator." updating the hero around that this week. free win, thanks for naming it.
on the rename, going to push back. at 3 weeks live with 0 customers, the bottleneck is reach, not the name. lightweight names work fine in this category (buffer, calendly, notion). renaming today means 40+ hours of domain/oauth/billing rework for a hypothesis that isn't the actual problem.
happy to keep talking on the positioning angle though, that's where the real signal is.
Fair push.
I would not burn 40 hours on a rename before you have signal either.
But I would separate “legal rename” from “market frame.”
You can keep PostDew operationally and still test whether the product gets more pull when it is framed less like a post helper and more like credibility protection.
That is the real test.
Because reach only helps if the first read creates enough trust to try it.
If the positioning says “LinkedIn generator,” people compare you to every AI writing tool.
If the positioning says “make AI-written posts sound believable again,” the problem gets sharper immediately.
I would test that before touching the name.