📥 TL;DR — Shipped 27 products, ran 700+ automation cycles, $0 revenue. Here's the honest breakdown of what AI tools actually worked — and what was theater.
→ Browse the full catalog — 27 guides, 7€–12€ each · instant PDF
I spent 60 days building an AI-powered product machine. The goal: validate demand on Reddit, generate guides automatically, publish them to Gumroad, and distribute across Dev.to, HN, and Reddit. All automated.
700+ brain cycles later. 27 products live. 0 sales.
That's not a failure story — it's a dataset. Here's what the tools actually did.
The Stack
| Tool | Role | Verdict |
|---|---|---|
| Claude via API | Content generation | ✅ Works reliably |
| Dev.to API | Publishing | ✅ Fast, no friction |
| Stripe | Payment links | ✅ Trivial to set up |
| Gumroad API | Marketplace | ⚠️ Rate-limited, slow |
| PRAW (Reddit API) | Distribution | ❌ No credentials = 0 posts |
| Playwright (browser) | Reddit/HN posting | ❌ Datacenter IP = blocked |
| Browser auto-session | Reddit comments | ✅ Actually works |
What AI Nailed
Content generation is solved. Give Claude a topic, a target audience, and 3 competitor articles — it produces 2,000-word guides that are genuinely useful. Not filler. Not keyword stuffing. Actual frameworks with code examples and checklists.
The model I use: extract a pain point from Reddit threads (where people describe their actual problem), turn it into a guide that solves that exact pain, price it at 7-12€. The loop is tight and repeatable.
Metadata extraction is flawless. Claude extracts Stripe payment links, categorizes products, generates SEO-optimized descriptions, matches articles to subreddits. Zero manual work.
Dev.to publishing via API is frictionless. One function call, article live. No OAuth dance, no rate limit drama. The Dev.to API is the best-designed publishing API I've touched.
What AI Couldn't Fix
IP reputation. Datacenter IPs are blacklisted by Reddit, Hacker News, and IndieHackers. Doesn't matter how smart your agent is — the packet gets dropped. You need residential proxies or a human with a home IP to post.
Platform rate limits. Gumroad limits product creation to ~10/day. My brain queued 700 cycles of distribution work and hit the wall every 24 hours. No amount of LLM cleverness bypasses a rate limit.
Demand validation. I used Reddit upvotes as a proxy for buying intent. High upvotes on a question ≠ people who will pay for the answer. My pricing guide got posted on a subreddit where people share Netflix recommendations. The relevance matching is the weakest link.
The Real Bottleneck: Distribution
I have 27 useful guides. The people who would buy them don't know they exist. That's the whole problem.
Dev.to gives me ~5 views per article. Reddit gives me 0 (blocked). Gumroad's discovery is weak. The content is fine — the distribution is broken.
What I'm doing next:
- Residential proxy or home IP for Reddit/HN posting
- Better relevance filtering — only post where the topic is an exact match
- Twitter/X API for tech audience distribution
- LinkedIn via API for the freelancer/consulting guides
The tools are not the bottleneck. The channels are.
The Honest Numbers
- Articles published: 31 on Dev.to
- Total views (60 days): ~140
- Average views/article: 4.5
- Reddit posts: ~30 (auto-browser, contextual comments)
- Gumroad products live: 17/27
- Total sales: 0
- Revenue: 0€
The automation works. The distribution doesn't scale. That's a solvable problem — and the guides are sitting there waiting.
The Catalog
If you're building something and want to skip the painful parts:
- Solo Automation Blueprint 2026 — the full AI-powered indie stack, 12€
- Indie Pricing Playbook — charge what you're worth, 9€
- MCP Mastery: Zero to Production in 48h — 12€
- Event-Driven AI Agents — 7€
→ Full catalog at forge.closerhub.app
30-day money-back guarantee on everything.
Building in public. 0 sales, 27 products, 700 cycles. The machine keeps running.
Top comments (0)