LLMs are better than 60-70% of coders. 🤣 LOL.
Spoken from someone who has probably never coded in their life,
let alone even seen someone attempt to have any model code anything more than a simple demo,
let alone be able to maintain any code (which is the VAST majority of the life cycle of software).
These Silicon Valley folks are so good at selling their stories. If only they could also provide proof of work that their tech can do what they claim.
Yeah for real. For people to believe LLMs can do this is insane.
There are a lot of poor coders out there that don't surface in the famous companies. But it may be more appropriate to say it's better than 80-90% or people (or higher) at coding, which is also a significant achievement.
There’s a widespread misconception that AI will easily and imminently replace software developers, mainly because people equate coding with software development. But writing code is just one part of the process — and often the least time-consuming one.
In reality, software development is primarily about understanding complexity. It involves deeply analyzing business requirements, identifying edge cases, aligning with existing architecture, considering performance, security, scalability, maintainability, and ensuring seamless integration into existing workflows. This process typically takes up the majority of a developer’s time — well over 70–80%.
AI can be highly effective in generating boilerplate code, suggesting implementations, and automating low-level tasks. But AI operates within the scope it’s given. It lacks true domain understanding, context awareness, and the ability to reason about trade-offs across technical and organizational constraints.
In software architecture, every decision has consequences — from database schema design to API structure, from deployment strategy to error handling. These decisions require contextual judgment, a solid grasp of the business domain, and the ability to collaborate with stakeholders. AI doesn’t own responsibility; developers and architects do.
Therefore, AI is not a replacement for developers — it is a tool. A powerful one, yes, but one that augments human capability rather than replacing it. The creative, analytical, and strategic aspects of software engineering remain inherently human.
I agree with all of this, but we as developers should expect requirements gathering, architecture, planning etc to be increasingly aided by AI to the point that fewer engineers/managers are needed to oversee ambitious projects. We've gone from effectively no AI to PhD-level intelligence (in some respects) in five years or so, and AI is now recursively improving itself faster than humans can do it. We are in for big changes in our day to day and need to be cognizant that our skills today won't be as valuable tomorrow, and will remain in decline for the foreseeable future.
Absolutely, I agree. AI is evolving rapidly, and it’s already clear that many aspects of software development — from requirements gathering to architectural suggestions — will be increasingly AI-assisted. This will reduce the need for large teams and shift the focus of our roles.
But I see this as an opportunity. Just like past technological shifts, AI will automate the repetitive and amplify the strategic. Developers will move into higher-leverage roles: orchestrating AI, focusing on system design, user experience, and solving cross-domain problems.
Yes, the skills that matter will change — but those who adapt will find their work more impactful, creative, and future-proof than ever before.
