DEV Community

Cover image for I Built a "Second Brain" Using — GitHub Copilot CLI
Lokesh
Lokesh

Posted on

I Built a "Second Brain" Using — GitHub Copilot CLI

GitHub Copilot CLI Challenge Submission

This is a submission for the GitHub Copilot CLI Challenge


The Moment Everything Changed

It was 2 AM.

I had 14 browser tabs open. A Notion doc I hadn't touched in weeks. A task manager I'd abandoned. A journal I forgot existed. A goal I set in January that was already dying quietly.

I was drowning in productivity tools that were supposed to save me.

So I opened my terminal, typed one sentence to GitHub Copilot CLI, and said:

"Help me build one app that replaces all of this."

What happened next changed how I think about building software forever.


What I Built

CorteXia — an AI-powered life operating system. Your second brain. One platform that actually understands your life.

Not another task manager. Not another habit tracker. Something deeper.

CorteXia connects everything:

  • ✅ Smart task management with goal linking
  • 🔥 GitHub-style 365-day habit heatmap
  • 🎯 Hierarchical goals (Year → Quarter → Month → Week)
  • 📅 Visual time blocking with drag-and-drop
  • ⏱️ Pomodoro focus sessions with analytics
  • 💰 Expense tracking with goal impact analysis
  • ✍️ AI-powered journal with mood tracking
  • 🤖 An omnipotent AI assistant that controls all of it through natural language

Complete a task? Your calendar updates. Hit a habit streak? Your goal progress moves. Tell the AI "create a high-priority task to finish my ML assignment by Friday 2pm"? It creates the task, blocks the time, and sets the reminder — in one sentence.

Tech stack: Next.js 14 · TypeScript · MongoDB Atlas · NextAuth · Tailwind CSS · Framer Motion

🔗 Live Demo
⭐ Star the repo: github.com/itslokeshx/CorteXia


Here's what the workflow looks like in real life:

Me: "I have a presentation Friday at 3pm, help me prepare"

CorteXia AI:
✅ Created task: "Prepare presentation slides" (High priority, Fri 3pm)
📅 Blocked: Thursday 7-9pm (Deep Work: Prep)
📅 Blocked: Friday 1-2:30pm (Final Review)
⏰ Reminder: Friday 12:30pm
🎯 Linked to goal: "Q1 Career Growth"
Enter fullscreen mode Exit fullscreen mode

One sentence. Six actions. Zero context-switching.


My Experience with GitHub Copilot CLI

Here's the honest truth: I couldn't have built this alone in the time I did.

The Turning Point

I started CorteXia as a weekend experiment. By Sunday night, I had a basic task manager. Nothing special.

Then I started using GitHub Copilot CLI differently.

Instead of asking it to write code, I started having conversations with it.

gh copilot suggest "how do i structure a mongodb schema 
where tasks, habits, and goals are all interconnected 
and can reference each other bidirectionally"
Enter fullscreen mode Exit fullscreen mode

It didn't just give me a schema. It explained the tradeoffs. Denormalized vs normalized. When to embed vs reference. What indexes I'd need at scale.

I went from "I think this might work" to "I understand exactly why this works."

The Debugging Superpower

The moment I became a believer:

I had a race condition in my task completion toggle. Tasks were sometimes unticking themselves. Duplicating. It was subtle, intermittent, and maddening.

I described the bug in plain English to Copilot CLI:

gh copilot explain "my react optimistic update is causing 
duplicate state entries when the user clicks the checkbox 
quickly twice before the api responds"
Enter fullscreen mode Exit fullscreen mode

In 30 seconds, it identified the exact problem — I wasn't debouncing the handler AND my cache invalidation was firing before the mutation settled. It gave me the fix. It explained why the fix worked.

What would have taken me 2 hours of console.log debugging took 4 minutes.

Building the AI System

The most complex part of CorteXia is the AI assistant — a system that:

  • Builds real-time context from your live data
  • Understands multi-action commands
  • Executes database operations from natural language
  • Rotates between API keys automatically

I used Copilot CLI to architect every layer:

gh copilot suggest "design a prompt engineering system 
that injects live user data (tasks, habits, goals) 
into an LLM context window without hitting token limits"
Enter fullscreen mode Exit fullscreen mode

The result was a context-compression algorithm I never would have designed myself. It prioritizes the most relevant data based on what the user is asking. Asking about tasks? It loads your last 10 tasks and today's time blocks. Asking about goals? It loads progress data and recent completions.

The AI became smarter because Copilot CLI made me smarter.

What Surprised Me Most

I expected Copilot CLI to be an autocomplete on steroids.

What I got was a thinking partner.

When I was stuck on how to handle offline sync across the app, I didn't Google it. I just asked:

gh copilot suggest "best pattern for queuing failed api 
calls in a next.js app and replaying them when the 
connection is restored"
Enter fullscreen mode Exit fullscreen mode

It walked me through the exact pattern. I implemented it in 45 minutes.

When I needed to optimize MongoDB queries that were slowing down the dashboard, I pasted the query and asked why it was slow. It explained compound index strategy in plain English and rewrote the query with proper index hints.

The CLI became my rubber duck — except it talks back with the right answers.


The Bigger Lesson

We talk a lot about AI replacing developers.

After building CorteXia with GitHub Copilot CLI, I believe the opposite.

It didn't replace my decisions. It accelerated them.

I still designed the architecture. I still made the product decisions. I still wrote the logic. But every time I hit a wall — a concept I didn't fully understand, a bug I couldn't see, a pattern I'd never implemented — Copilot CLI collapsed the gap between where I was and where I needed to be.

CorteXia went from idea to full-stack production app with 12 database collections, an AI assistant, 8 feature modules, and a world-class UI — faster than any project I've ever shipped.

That's not AI replacing a developer.

That's a developer with a superpower.


Built with late nights, strong coffee, and GitHub Copilot CLI. If this resonated with you, drop a ❤️ — it means more than you know.

Top comments (3)

Collapse
 
anmolbaranwal profile image
Anmol Baranwal

this is cool. don't take this the wrong way but I think this post could be a lot better at least in terms of writing. how long did it take you to build this?

(I did try lol)

Collapse
 
itslokeshx profile image
Lokesh

🤝 i have been thinking about this product for months but haven't took the initiative to build this, after I saw the co-pilot challenge when pushed my limits to build it, where it took more than week to make it happen!

When made this post I was fully burnted out , thanks for your honest suggestion where I will drop a better writing for this!

Collapse
 
anmolbaranwal profile image
Anmol Baranwal

great work. I don't really mind the writing style (since everyone has their own preference) but I definitely think if you present this in a better way -- this might get a lot of traffic and positive feedback from the community.