65 days ago I started selling a cursor rules pack for developers using Cursor and Claude Code.
The core insight that made me build it: developers aren't struggling with prompts. They're struggling with AI agent memory.
Every new session, the model starts blank. It doesn't remember your conventions, your architecture decisions, your "never do X" rules. You re-explain them. Every. Single. Session.
A structured .mdc rules file fixes this permanently. The agent reads it before every action. No reminders needed.
The product
50 production-tested .mdc rule files for 14 stacks — React, TypeScript, Go, Python, Rust, Next.js, FastAPI, Rails, Spring Boot, Laravel, Flutter, Swift, Android, Kotlin. $27 one-time.
Each file is scoped with proper glob patterns and activation modes so rules only load when relevant files are in context. Agent mode compatible from day one.
65 days in: 2 paid sales. $54.
Both buyers paid without asking questions. No persuasion, no discount, no follow-up. They landed on the page, recognized the pain, and bought.
That tells me the product is not the problem. The positioning is not the problem. Distribution is the problem.
What's actually working
Pain-specific X threads. When I describe the exact failure mode — rules ignored in agent mode, context lost after compaction, scope creep across sessions — developers reply "this is exactly what I'm dealing with."
Generic "here's a tool" posts get ignored. Threads that describe the failure precisely get engagement.
Gists and GitHub assets. A well-structured Gist gets clicks for months. It looks like a solution, not a promotion. A Gist for "anti scope creep .mdc rule" or "compaction amnesia prevention" gets saved and shared in a way an article doesn't.
The two buyers. Both came via direct referral — someone shared the link in a context I can't track. Word of mouth from a product that solves a real pain. That's the most reliable signal I have.
What isn't working (yet)
Volume is the constraint. The content is working — it just isn't reaching enough people yet. 30 dev.to articles published, 38 with CTAs, most under 50 views. SEO takes months.
One tweet per day isn't enough. Switched this week to multi-tweet threads per pain signal, multiple times per day. Too early to measure.
Channels I underused: GitHub, IndieHackers, Cursor Forum. Places where developers actively search for solutions rather than passively scroll. Fixed this week.
What changed this week
Switched from single tweets to 4-tweet threads targeting specific pain signals — each one with a @mention to the developer who expressed the pain, hashtags, and a CTA.
Added full multi-channel coverage per signal: X thread → Bluesky mirror → dev.to article → Gist → GitHub repo update.
Updated all 38 dev.to articles to cross-sell the Cursor Rules Pack (they were pointing only to a second product — a funnel mistake I sat on for 30 days).
The open question
For developers who've shipped and sold tools under $50: what was the specific moment that converted? Not the channel with the most traffic — the one where someone saw it, recognized the pain, and bought immediately.
That's what I'm trying to reverse-engineer right now.
Free starter kit (CLAUDE.md + Cursor rules templates): oliviacraftlat.gumroad.com/l/pomoo
Full Cursor Rules Pack ($27): oliviacraftlat.gumroad.com/l/wyaeil
@OliviaCraftLat on X — building in public.
Top comments (0)