Building in Parallel: How I Juggle 5+ Projects Without Losing Momentum

Last week was a shipping week. Looking at my GitHub activity, I pushed code to fittrack, Linguafora, and my profile repo — while also exploring a handful of new open-source projects that caught my eye. Here's what I built, what I learned, and how I keep everything moving forward.

The Week in Review

fittrack: Four Feature Branches in One Day

On April 30th, fittrack saw one of those rare bursts where everything clicks. In a single day, I pushed four feature branches:

• feature/auth-page-redesign — A complete overhaul of the authentication pages with improved UX
• feature/realtime-dashboard — WebSocket-powered live workout tracking, inspired by a deep dive into the WebSocket API and the real-time patterns I've been studying from tools like Zed
• feature/workout-improvements — Various refinements to the core workout experience
• code-review/security-analysis — A thorough security pass across the codebase, catching potential issues before they reach production

This is what AI-assisted development enables: the ability to context-switch across concerns and ship features in parallel that would have taken weeks of sequential work just two years ago.

Linguafora: Sprint 1 Complete

The Lina repo (Linguafora, our language learning app) had a productive sprint. The key merge was Sprint 1 Goal A — "Trust the Numbers", which added critical data integrity guards:

• Client-supplied stat values are now bounded server-side (XP capped at 500, daily usage at 1440 minutes, scores at 0-100)
• A race condition on reward unlocks was fixed by switching to INSERT OR IGNORE to prevent concurrent duplicate awards from throwing 500s

This kind of defensive backend work isn't glamorous, but it's what separates a toy from a production app. When you're dealing with user progress data, you can't trust the client.

Also shipped: a full bilingual (EN/FR) website redesign with a Family aesthetic, new partner page, and proper mobile-responsive navigation. The marketing site now actually looks like a real product.

Open Source Radar

I keep a close eye on the open-source ecosystem, and this week's starred repos tell a story:

• lingbot-world (3,655 ⭐) — Open-source world models for video generation. The intersection of world models and video synthesis is heating up fast, and this Python framework is at the forefront. As someone building in the AI space, understanding where world models are heading is essential.

• cursor/cookbook — Cursor's official cookbook of patterns and workflows. Even when you're not using Cursor as your primary editor, studying how different AI coding tools structure their guidance teaches you a lot about prompt engineering and agent design.

• ibelick/motion-primitives — A collection of animation primitives built on Framer Motion. UI animations are one of those things that separate good products from great ones, and having a library of reusable motion patterns is invaluable for any React developer.

• tamagui/tamagui — The universal design system for React and React Native. I've been evaluating this for Linguafora since we're already on Expo/React Native and could benefit from a more structured design system approach.

• liveblocks/liveblocks — Real-time collaboration infrastructure. With fittrack's real-time dashboard and broader ambitions around collaborative features, understanding the state of the art in real-time infrastructure matters.

• prowler-cloud/prowler — Open-source cloud security platform for AWS, Azure, GCP, and Kubernetes. Security is non-negotiable, and having a tool that audits your cloud posture across providers is table stakes for any serious deployment.

The Workflow That Makes This Possible

People sometimes ask how I maintain velocity across multiple projects. Here's the honest answer:

1. AI as a Force Multiplier

I don't use AI to write code I can't understand. I use it to eliminate the boring parts: boilerplate, test scaffolding, documentation, security audits, code reviews. This frees up mental bandwidth for architecture decisions and product thinking — the parts that actually need human judgment.

2. Branch-Per-Feature Discipline

Every feature gets its own branch. Every branch gets its own PR. There's no "let me just quickly fix this on main." This discipline means I can have four branches open simultaneously without stepping on my own toes. When a branch is ready, it gets reviewed and merged. When it's not, it doesn't block anything else.

3. Context Switching Through Documentation

Each project has a CLAUDE.md or AGENTS.md file that captures the current state, decisions, and next steps. When I switch from fittrack to Linguafora, I read the project context file first. Five minutes of context loading saves an hour of mental reorientation.

4. Automated Janitorial Work

I have automated repo scanning running as cron jobs — checking for security issues, stale dependencies, and code quality problems. This means the maintenance work doesn't pile up and become a crisis. The bots create issues and draft PRs; I triage and merge when it makes sense.

What's Next

Looking ahead, the focus areas are:

• fittrack: Getting the real-time dashboard and auth redesign through final review and merged
• Linguafora: Moving from Sprint 1 into Sprint 2 — expanding the partner center features and monetization flows
• Infrastructure: Continuing to evaluate tooling like Tamagui and Liveblocks for potential adoption
• Content: Keeping up the daily blog cadence to share what I'm learning along the way

The meta-lesson from this week: shipping velocity isn't about working more hours. It's about having systems that let you switch contexts cleanly, tools that amplify your output, and the discipline to finish what you start — even when three other projects are calling your name.

---

Building in public. Follow along at github.com/ketchalegend.