CoLeader is a personal development and founder-leadership app with 10 archetypes, a working Leadership Journal, an auto-generated Action Plan system, and a Wisdom tab with 20 real leadership quotes and 20 glossary concepts. It took weeks of building, a brutal production debugging session, and more EAS builds than I care to count. Here’s everything I learned.
The Product Idea
CoLeader started as a simple question: what if the CoFamily had an app for people who feel like they’re meant to build, lead, or contribute more — but don’t know what to do next?
The CoFamily already had CoFun for laughs, CoWiz for reflection, CoVerse for creative expression. CoLeader was going to be the one that said: okay, now what are you actually going to build with all this?
The core experience: pick a situation, pick a leadership archetype — Visionary Founder, Startup Operator, Mastery Coach, Purpose Communicator, and seven more — describe what you’re working through in your own words, and get a structured response that moves you forward. Not generic motivation. A real perspective from a specific leadership lens.
“Sometimes growth comes from being deeply understood. Sometimes growth comes from being seen fresh. CoLeader is built around that second idea — every session starts clean, no history, no assumptions, no excuses.”
— From the CoLeader LinkedIn launch post
The Architecture
Every CoFamily app is built on the same fork-and-theme engine. The three swappable layers per app are: voice/archetype data and system prompts, bit/situation data, and visual design tokens. Everything else — navigation, generation flow, thread system, freemium logic, save and share — is shared infrastructure.
React Native Expo
SDK 54, Expo Router, EAS Build
Anthropic Claude API
claude-sonnet-4, Haiku for secondary calls
Cloudflare Workers
API proxy with secret header auth
RevenueCat
Google Play billing, entitlement management
AsyncStorage
Local state, journal, preferences
Google Play Console
Internal → Closed → Production pipeline
New Features I Built For This App
CoLeader introduced several features that didn’t exist in previous CoFamily apps and will now propagate back across the family.
New
Leadership Journal
The Working Reflect Screen + Leadership Journal
The biggest UX departure from CoWiz. Instead of a conversation screen that resets every session, CoLeader has a Reflect screen that stays locked to the current session until the user explicitly taps “Start Fresh.” The full session — original response, all Another Takes, the Action Plan — saves together as a named journal entry.
Users can name each session (“my app idea,” “the promotion conversation”), add multiple timestamped notes, and come back to it. The journal feels like a leadership log, not a chat history.
New
AI-Powered
Auto-Generated Action Plan
After generating a response the user can tap “Create My Action Plan.” A secondary Haiku API call extracts 3–5 concrete, specific actions from the full session — original response plus all Another Takes. The checklist persists with checkbox state and saves to the journal entry.
The system prompt for the extraction call instructs the model to return only a JSON array of strings — no preamble, no markdown, no fences. This is the pattern I use for all structured secondary calls across the CoFamily now.
New
Multi-Take
Another Take with Archetype Switching
Up to 3 Another Takes per session. Each take includes an archetype picker — the user can keep their current archetype or switch to a completely different one. A faith-based startup idea might get the Startup Operator take, then the Servant Leader take, then the Moral Courage Leader take. Three completely different lenses on the same situation.
Each take saves to the journal with its archetype badge clearly labeled.
New
Content
Wisdom Tab — Real Quotes + Real Glossary
20 real leadership quotes from real leaders — JFK, Eleanor Roosevelt, Simon Sinek, John C. Maxwell — filterable by theme. 20 real leadership concepts — Psychological Safety, The Multiplier Effect, First Principles Thinking, Radical Candor — with plain-English definitions.
A daily rotating quote banner on the Home screen pulls from the same data, cycling through all 20 based on the date. Tapping it navigates to the full Wisdom tab.
What I Learned — The Technical Lessons
01
Always open a new PowerShell and cd to the correct project before any Claude Code session
During CoLeader development, Claude Code accidentally wrote 13 files into the CoFun project directory. The changes had to be fully reverted from CoFun and reapplied to CoLeader. This has now happened across multiple apps. The fix is discipline — always verify the working directory before any session starts.
02
versionCode continuity matters more than you think
CoLeader’s production launch was delayed by multiple days because early accidental builds from the CoWiz project consumed versionCodes 1–5 in Google Play Console. When the real build was submitted at versionCode 2, Google Play silently kept the old broken build. The fix was jumping to versionCode 6. Always track what versionCodes have been consumed across all tracks — internal, closed, and production.
03
A 15-second fetch timeout is too short for the Anthropic API on mobile
The most expensive debugging session in this project was caused by a Promise.race timeout wrapper set to 15 seconds. The Anthropic API with max_tokens at 1500 regularly takes 20–30 seconds on a mobile network. The app worked perfectly in Expo Go (faster dev connection) but silently timed out on every production build. The fix was either removing the timeout entirely or setting it to 60 seconds. Never add a timeout shorter than 45 seconds on any CoFamily API call.
04
The X-CoFun-Secret header must be in every new app’s callProxy from day one
CoWiz, CoVerse, and CoLeader all shipped production builds at some point without the Cloudflare Worker secret header, causing silent 401 rejections. The header is now explicitly checked in every pre-build audit. It lives in a single line and it has caused more hours of debugging than any other single issue across the CoFamily.
05
The Cloudflare Worker abuse detector will block your own IP during intensive debugging
The abuse detection system blocks IPs that send more than 20 requests per minute. During a heavy debugging session, this threshold was exceeded and the home network IP was blocked for an hour. The limit has since been raised to 60. Any IP that’s actively developing should be excluded or the limit set high enough that a developer running test builds doesn’t trigger it.
06
Google Play App Signing solved the keystore mismatch problem permanently
After an accidental build from the wrong EAS project registered a CoWiz keystore fingerprint against the CoLeader Play Console listing, the app couldn’t be updated. The fix was switching to Google Play App Signing — Google now manages the distribution key entirely. EAS upload keys are separate and Google re-signs the AAB before distribution. This is now the standard approach for all new CoFamily apps.
07
RevenueCat entitlement key must match the dashboard exactly — every character
The entitlement key in code must match the RevenueCat dashboard string with exact capitalization and spacing. “CoLeader Pro” in the dashboard and “coleader_pro” in code are different things. This mismatch caused a silent billing bug in CoWiz earlier in the project and was caught only because purchases were completing but Pro wasn’t unlocking. Always verify the exact string before any production build.
08
When a production build fails mysteriously, match the failing code to a known-working implementation
After exhausting every diagnostic approach, the fix that actually worked was replacing CoLeader’s callProxy function with a byte-for-byte copy of the working CoWiz implementation. Same URL, same headers, same response parsing, same max_tokens. Analytical debugging had been unable to find the difference. Copying the working code found it immediately — a subtle structural difference in how the fetch response was being parsed.
What I Learned — The Product Lessons
01
A working screen beats a conversation screen for goal-oriented apps
CoWiz and CoFun are conversation apps — generate, follow up, generate. CoLeader needed something different. The Reflect screen as a persistent working space, stable until explicitly cleared, changes how users interact with the content. They come back to it. They add notes. They check off action items. The session has weight. That’s a fundamentally different product feel.
02
Starting fresh can be a feature, not a limitation
Most AI tools are getting better at remembering you. CoLeader sessions start clean — no history, no assumptions, no prior context. This turned out to be a genuine differentiator. Users get perspective from a lens that has never heard their story before. Sometimes growth comes from being deeply understood. Sometimes it comes from being seen fresh. CoLeader is the second kind.
03
The free tier needs to demonstrate the best thing the app does
CoLeader launched with all 10 archetypes unlocked for free users. Only session count and Long Answer responses are gated. This was a deliberate choice based on zero early sales across the CoFamily — users need to experience the full quality of the product before they’re asked to pay. One Long Answer per day free gives users a genuine taste of the 7-part structured format. That’s the hook.
04
The frameworks I studied in grad school are invisible inside the product — and that’s intentional
Gilbert’s BEM, Kirkpatrick, ISPI HPT Model, Schunk’s learning theories — they informed every archetype, every output structure, every bit category. But users never see a framework label. The 7-part output format (Leadership Reframe, Hard Truth, Next Brave Move, Words You Can Say, 10x Habit, Contribution Check, Reflection Question) is instructional design dressed up as leadership coaching. That’s the differentiator that competitors can’t easily replicate.
The Build Timeline
📋
Product Brief + Claude Code Brief
Full product strategy, 10 archetype definitions, bit library, screen flows, freemium model, and design tokens documented in a 10-section Claude Code brief before any code was written.
🔧
Initial Fork from CoWiz
Theme swap, content layer replacement, 3-step input flow (bit → archetype → situation text), output length pills, thread continuity implementation.
✨
Feature Expansion
Leadership Journal, Action Plan builder, Another Take with archetype switching, Wisdom tab with real quotes and glossary, Continue Where I Left Off banner, Named Sessions, light/dark theme toggle.
🧪
14-Day Closed Test
Feedback collected via written report. Key fixes: bottom nav position shift on load, favourited quotes page added to Profile, settings section restructured for clarity.
💳
RevenueCat + Google Play Billing
Production subscription wired: coleader_pro_monthly at $4.99/month, base plan coleader-pro-monthly-base, entitlement CoLeader Pro, Google Play App Signing activated.
🚀
Production Launch
14 production builds, a Cloudflare abuse block, a versionCode collision, a fetch timeout bug, and one very long debugging session later — CoLeader went live on Google Play.
What This Project Means to Me
CoLeader is the fourth app I’ve shipped to Google Play as a solo developer. Each one has taught me something the last one didn’t. CoFun taught me the CoFamily architecture. CoWiz taught me RevenueCat. CoVerse taught me that the fork discipline matters. CoLeader taught me that production and development are genuinely different environments and you have to respect that difference.
But the lesson I keep coming back to is simpler than any of that. I have an MS in Learning Design and Technology from Purdue. I spent years studying how people learn, why performance changes, what actually moves behavior. CoLeader is the first consumer app where I feel like those years of study are genuinely inside the product — not as jargon, not as academic citation, but as invisible architecture that makes the output better.
That’s what I’m building toward. AI apps with real intellectual foundations that users experience as personality and quality, not theory.
Lead yourself first. Then build something that matters.
Leave a Reply
You must be logged in to post a comment.