A user recounts a conversation where their retired CTO uncle argues AI will replace developers, but the user counters that human validation, cost, and risk management still require human oversight, dismissing the idea as overhyped.
So i was talking to my uncle last night who is a retired CTO and said Microsoft created an AI test harness that will take code (AI generated or not), search for vulnerabilities, fix them and then provide an overview of all the changes. I thought sounds great on paper, but we still need validation that it did the job right. He then looked at me and said "why would we need to validate if in the future models are getting better. Im just not sure where devs will fit in the world anymore." But I thought going back to the original test harness, the AI checking for vulnerabilities still needs code, so if its generate by AI is it not almost like checking its own homework, right? Then were not considering cost of resources, which granted will get better over the next few decades (we hope) to house better models but will it truly have human level reasoning? It doesnt gel with me that the entire process of product creation, testing and validation is all done via LLMs and then straight to production (cause AI can now build IAC now eliminating the need for cloud engineers aswell according to him). This entire take sounds ok on paper for anyone with a tech business or a few million to invest but when you actually use a little bit of non-AI influenced brain power i can think of so many things going wrong. Token cost running a business tech/IT budget to zero, production destroying bugs and then the non-existing devs having no idea what the code does, then IAC being incorrect could absolutely destroy the auto-scaling and slowly ramp up the cost without that level of validation and fine-tuning. What does this community think? Personally ..... I think my uncle is on the AI overhyped train. Edit: I also would like to say it did say most of this to him and he said well that why you properly plan and need to create an extremely details prompt with specific rules and edge cases so it captures it all.... that just sounds like coding in plain English to me! But with more bugs and more cost!
The article discusses Microsoft's policy against employees using AI for code and argues that the rapidly decreasing cost and increasing speed of AI will make it difficult for human developers to compete, challenging the idea that AI won't replace developers.
Arvind Narayanan and Sayash Kapoor argue that AI will not replace software engineers, citing bottlenecks like deciding what to build, verification, and deep human understanding as key to the profession's resilience against automation.
The article argues that the real shift from AI is not job replacement but a change in who makes decisions, emphasizing the growing importance of human judgment and supervision over AI systems.
Cognition CEO Scott Wu discusses that AI coding agents like Devin are designed to assist, not replace, human programmers, emphasizing human-AI collaboration over job displacement.