Table of contents

How Agentic Coding Transformed My Junior Dev Workflow in Just 2 Days

Daylee    |    2026-05-08

Changing Perspective

Recently, I built a LinkedIn job filtering app that saves me two hours a day. It took six iterations to build this bespoke software with the help of an AI agent.

Previously, I had held myself back from using coding assistants like Claude and GitHub Copilot. As a junior developer, I wanted to build my “coding muscle.” I believed that a solid foundation and true craftsmanship could only be developed through manual trial and error. However, I soon realised I was being a bit of a “caveman.”

“It’s just another tool.”

My perspective shifted after attending the Monki Gras conference in London. I had the chance to talk to engineers with over 25 years of experience, people whose careers evolved alongside Docker and AWS. When I shared my hesitation toward agentic coding, they all said the same thing: “It’s worth building something with an agent; it’s just another tool.” The landscape is shifting rapidly. A few months ago, companies were worried about the security risks of AI. Now, many are mandating the use of AI agents for efficiency.

The Concept of the “Single-user App”

The idea of the “Single-user App” further motivated me to build personal software. As Betty Junod from Monki Gras put it: “The AI app dev movement will see an explosion of highly custom apps designed by and for a single user."

I was always told to build apps to hone my skills, but spending significant time on a side project that only catered to me never felt time-efficient. I have many forgotten “passion projects” because I felt discouraged that they weren’t “scalable.” The concept of a custom, single-user app, built to solve my specific problems via “vibe coding” finally justified building something just for myself.

How a Junior Developer Approached AI-assisted Coding

A LinkedIn job filtering app

Initially, I planned to build my own scraper. However, I learned that LinkedIn is strict about scraping, and the risk of being banned was too high. Instead, I used Apify, a professional scraping service. It was cost-effective and, most importantly, safe. With an AI agent, I built an automated system to filter and save that data into Google Sheets.

linkedin custom filter app

Linkedin Custom Filtering App

The Agentic Coding Journey with GitHub Copilot:

  1. Context Setting: I explained the project goals, design, and requirements clearly.
  2. Prompt Engineering: I asked how to write better prompts for the Agent to minimise bugs from the start.
  3. Iterative Refinement: We refined the requirements until every detail made sense to both me and the AI.
  4. Modular Code Generation: I broke the work into five small phases. I reviewed the code after each phase rather than doing it all at once.
  5. Data Verification: I added tests and cross-checking code to verify the raw data.
  6. User Testing: I tested the app myself to find bugs and edge cases

Note: I later learned the importance of setting a copilot-instructions.md file for context. Previously, I was manually pasting context for every phase.

linkedin custom filter app

Agentic coding

I didn’t just perform “vibe coding” (building purely with natural language). Instead, I practiced AI-assisted coding, treating the AI as a true pair programmer. I actively reviewed code, made manual tweaks, and maintained a spec document. The project was built in gradual phases under my careful monitoring.

Agentic coding is an unprecedented and inevitable trend. As a junior developer navigating a tough job market, the only way forward is to adapt to this ever-changing industry. We must pivot our skills just as veteran engineers have evolved their toolsets throughout their careers.