📝 Content Roadmap

Ideas → Drafts → X threads → Published. Living document, updated as projects evolve.

📅 Updated May 12, 2026 📦 8 ideas 🚀 1 posted
Posted Published on X Idea Captured, not written yet ☑️ Picked for X
🚀 Published
Live content on X. Posted from this roadmap.
✓
Small Pain Points Deserve Small Price Tags Posted
#Philosophy 6 tweets 📅 May 12, 2026

The thesis is live. First thread posted on X from @RuidiYuan. Establishes the core philosophy everything else builds on.

📖 View thread draft

How X suitability is rated

X ★ Great — Has a built-in narrative arc. A clear hook in tweet 1, a middle that builds tension or reveals pattern, and a satisfying conclusion. Can be told in 5-7 tweets without losing depth. The reader walks away thinking "huh, that's interesting" without needing prior context. Examples: debugging stories (OAuth bug), opinionated takes (small pain points), personal observations (MVP on a plane).

X OK — Has some thread potential but one of: (a) the interesting part is a single insight, not a narrative, (b) it needs setup/context that makes tweet 1 too long, or (c) the examples/data are what make it good, and those work better formatted. Still could work as a thread with the right angle.

Long-form — The idea needs space. User quotes, data tables, architectural diagrams, or nuanced arguments that you can't compress into tweets without losing what makes it interesting. Start as Substack. You can always excerpt the best part for X later.

Phase 1: Establish the Lens
Start with the broad philosophy — set up how I think about building things. Then follow with a concrete story that proves it.
✓
Small Pain Points Deserve Small Price Tags
Phase 1 #Philosophy 🚀 Posted May 12

✅ Posted to X. This thread established the thesis: small problems deserve small price tags.

📖 View the thread draft

2
Why Users Didn't Want AI Chat for Their Inbox Idea
Phase 1 #Building Long-form
☐

Thesis: Users didn't want a chat interface for their email. They wanted a digest. This UX research (TED-38) revealed that "AI polish" (natural language → structured preview) beats "AI chat" every time for utility tasks.

  • Users described chat as "another thing to manage" — they want less inbox, not another inbox
  • The three-state category modes UX research showed users prefer implicit structuring over explicit commands
  • Analogy: the Roomba vs the robot butler. Roomba works silently in the background; the butler requires conversation
  • Good for Substack — needs data, user quotes, and room to breathe
Phase 2: Depth (3-4 posts)
Once the lens is established, go deeper into specific builds — technical memoirs and process stories.
3
The Google OAuth Bug That Almost Broke My Pipeline Idea
Phase 2 #Building X ★ Great
☐

Thesis: A specific, painful debugging story with a satisfying resolution. The kind of post that every developer who's fought OAuth will nod along to.

  • Concrete bug that took hours to diagnose
  • Teachable moment about OAuth refresh tokens, scopes, or consent screen quirks
  • Has a built-in narrative arc: problem → investigation → discovery → fix
  • Perfect for X — 6-7 tweet thread with the "aha" moment as the climax
4
I Track Every Pageview Through a Cloud Function Idea
Phase 2 #Building Long-form
☐

Thesis: DIY analytics with PostHog + Firebase Cloud Function reverse proxy. A technical deep-dive into why the author chose self-hosted-analytics-over-a-proxy over Google Analytics — and the surprising things you learn when you own your data pipeline.

  • PostHog reverse proxy via Firebase Cloud Function — the architecture
  • Cookie-less tracking, privacy-first approach
  • Lessons from running your own analytics: what you see vs what GA shows you
  • Better for Substack — technical details benefit from long-form
5
What 50+ A/B Tests Taught Me About Users Idea
Phase 2 #Philosophy X OK
☐

Thesis: Cross-pollination from iAdvize days. The counterintuitive lessons from running 50+ A/B tests — not about which variant won, but about how users behave when they know they're being tested, and what "statistical significance" actually means in practice.

  • The test that "failed" but taught us more than the winners
  • Surprising user behaviors that no metric predicted
  • The gap between what users say and what they do
  • Could go either way — one strong insight works as an X thread; the full story needs Substack
6
Shipping a SaaS Alone Without Burning Out Idea
Phase 2 #Building X ★ Great
☐

Thesis: Solo SaaS development is a totally different game from team development. The bottleneck isn't code — it's energy, context-switching, and knowing when to stop. A personal take on pacing, delegation (to AI), and the discipline of "done."

  • Your bottleneck isn't coding speed — it's decision fatigue
  • How AI delegation changes the solo dev calculus (Hermes handles cron, research, monitoring)
  • The 80/20 rule on steroids: what you actually need to ship vs what you think you need
  • Great for X — relatable, conversational, easy to thread
Phase 3: Breadth (ongoing)
Lighter, more personal pieces. Mix of lifestyle observations and building stories.
7
The Best Time to Build an MVP Is on a Plane Idea
Phase 3 #Personal X ★ Great
☐

Thesis: There's something about the constraints of a plane — no internet, nowhere to go, limited battery — that forces the kind of focus you can't manufacture in daily life. The best MVPs get built when you have nothing else to do.

  • A personal anecdote about building something on a flight
  • The psychology of constrained environments vs open-ended ones
  • Why "I'll work on it this weekend" never works but "I have 3 hours with no wifi" does
  • Perfect X opener — hooky premise, relatable, short
8
Cash Flow Tracking in a Weekend Idea
Phase 3 #Building X OK
☐

Thesis: You don't need a budgeting app. You need a spreadsheet that knows your habits, or a tiny app that does one thing — track cash flow — without Mint-style feature creep. Another "small tool for a small problem" story.

  • The build: what stack, how long, what shortcuts
  • Reinforces the central thesis (small tools for small problems) with a concrete example
  • Good for Substack — builds on the philosophy established in post #1
9
How I Save Tokens by Delegating to the Right Model Idea
Phase 3 #Building X ★ Great
☐

Thesis: Most LLM queries don't need your daily driver. Three-tier delegation saves tokens without sacrificing quality. Bonus: even within the same model, different providers can have drastically different cache hit rates (82% vs 16%), which means the same output at 4.5× cheaper.

  • Tier 1 (free): research, brainstorming, non-sensitive — free model
  • Tier 2 (fast): daily coding, quick questions — main model
  • Tier 3 (powerful): architecture, complex debugging — best model
  • Provider routing: same model, different API, different cache rate, different price
  • Real data from my OpenRouter monitor (cron job tracks this daily)
  • Perfect for X — concrete tip with numbers, clear tiers, instantly useful
☑️ X Pick List — 0 selected
☑️

Click the checkbox on any card to mark it for X.

Your picks will show here as a prioritized list.