@tanujDE3180 Default answer should be the staff engg, because they have enough real world experience and technical depth to verify LLM output.
For junior enggs, it depends on mindset: are they using AI just to get things done, or are they actively trying to understand what the AI is doing?
Can't deploy publicly - Spotify's API now restricts solo developers from extended quota access
But it's fully functional locally. Clone and run it if you want to organize your own library
My friend creates a new Spotify playlist every month
I have 900 songs in one playlist and wanted that nostalgia - what I was listening to during specific periods
Built Monthlify to automate this
Demo: https://t.co/g63BaheLHs
Code: https://t.co/x487IqmVcH
Takes any Spotify playlist, groups songs by the month you added them, creates separate playlists with auto-generated covers
Built with Next.js + Flask. OAuth handles auth, Spotipy manages playlist operations, Pillow generates unique cover art for each month
@yourclouddude The real issue is these tests don’t reflect how engineers work now. I don’t support using AI code blindly, but AI is part of the job. A better screen would check how candidates use AI to reach a solution, then confirm understanding with a short call where they explain it.
@yourclouddude I agree with this path. I’ve never enjoyed LeetCode, and it feels even more disconnected now that AI can generate complex algorithms instantly. But big companies like Google or Meta still rely on DSA tests because they need a scalable way to filter thousands of applications.
@hasantoxr For repeated workflows, I usually write one comprehensive instruction set and save it in ChatGPT/Claude projects. It keeps the context consistent without having to re-enter the full prompt each time.
@hasantoxr ‘Context stack’ is interesting wording, but it feels like a cleaner, more structured version of what was already possible with a well-written prompt. New label, same underlying idea.