The Full Episode:

Spotify - https://open.spotify.com/episode/4pqIupaKYWHl8D4pZMr5hL?si=1-ESZEHzSTK8K_wJyxLFLw
Apple - https://podcasts.apple.com/us/podcast/building-the-future-of-coding-opencode-with-dax-raad/id1867041268?i=1000752897650


For sponsorship & business inquiries mail@neetcode.io


🚀 https://neetcode.io/ - A better way to prepare for technical interviews

Second Channel:  https://www.youtube.com/@NeetCodeIO
🧑‍💼 LinkedIn: https://www.linkedin.com/in/navdeep-singh-3aaa14161/
🐦 Twitter: https://twitter.com/neetcode1

#coding #leetcode #python

NeetCode

A developer pushes back on claims that writing code from scratch is dead. They argue that agentic AI coding tools aren't great for code review, and that manually typing code is itself part of the design thinking process — helping developers figure out function relationships and folder structure. They draw a parallel to past tech hype cycles where people claimed a new tool changed everything, but the resulting products often didn't exist or weren't better than before. The skepticism extends to products built entirely with AI, which often feel lower quality.

Is Writing Code From Scratch Dead?

Most AI models support planning or most devs already have their own context files to speed up prompting and weed out basic hallucinations. To write that context or to write a full plan is basically an architectural overview of the program which you have to know/understand before you create the code. Given this whole document, will ensure you already thought it through and itemized steps. Its somewhat easier to write these out in plain language than straight to program code. If you just start writing code you get caught up in implementation issues/granularity and don’t see if there are any big picture problems you will encounter. The AI can judge your plan and challenge it to well known progammatic practices, principles and already detect higher level issues. Also for the AI its way simpler to double check all work and refactor parts when you figure out there is something wrong with implementation or design. Its less likely to forget to change all related files when refactoring than your average dev. I hate how rigid people can be of moving forwards, I need to test and debug every functionality anyway. I can just have the AI write the code, I gloss over to see what are you doing and what is the idea, then debug. I might aswell just skip the coding part and focus on the thinking/designing/planning and debugging.

I think this is a healthy pushback. Writing code isn’t just about producing output, it’s a thinking tool. When you manually shape functions, name variables, and organize files, you’re actively modeling the system in your head. That cognitive loop is hard to replicate when you jump straight to AI generated solutions.
AI can help you move faster, but speed without clarity often leads to fragile systems. The real skill isn’t typing less, it’s understanding more. Developers who treat AI as a collaborator rather than a replacement will likely build better, more resilient software.

Doing everything by hand as if AI doesn’t exist anymore? I don’t think so. Even the most zealot of developers would at least use AI for the heavy lifting of scaffolding and bootstrapping a multi-actor system.

Well, I believe that blindly trusting all the code generated by AI is akin to continuously accumulating hidden technical debt.
Since AI tends to try to complete the task at hand as quickly as possible, the results aren’t always satisfactory. I believe that if you don’t develop a sense of attachment to and responsibility for your code, it will literally die.