Why True Software Developers Won’t Be Threatened by AI

Published on Medium: 2026-04-06 — https://medium.com/@danithaca/why-true-software-developers-wont-be-threatened-by-ai-8aa290f8acea

Before diving into this question, let me first explain why people feel threatened. My wife likes Brad Pitt, but I don’t feel threatened because nothing could realistically happen. Similarly, I play soccer casually, but I don’t feel threatened by Messi — only Ronaldo would see Messi as a threat. These examples illustrate a key point: we only feel threatened when someone (or something like AI) can take away what rightfully belongs to us.

Most people believe that a software developer’s job is to write code. Therefore, when AI can write code faster and better, it seems the software developer role would be greatly threatened. However, I believe that a true software developer’s core competence isn’t writing code — it’s being a builder who creates things from 0 to 1. In this building process, writing code is only a small part. What’s more important includes:

1. Deciding what to build

This is also described in Andrew Ng’s concept of the “product management bottleneck” (https://www.linkedin.com/posts/andrewyng_grok-raises-questions-meta-poaches-talent-activity-7351623555542896640-cTJT/). For example, when I’m building a note-taking app, should I support a tag cloud feature? AI cannot make this decision for me.

2. Deciding what to build first and what to build next

Building software is like constructing a house — you need to lay the foundation first, then build the first floor, then the second floor, then install doors and windows, etc. Everything must follow a sequence, and AI struggles to make these decisions at the architectural level.

3. Making opinionated tech design decisions

For instance, should we use Next.js or React? Should we use Anthropic’s Batch API? What about scalability and security considerations? AI cannot make these decisions well.

4. Code aesthetics

AI doesn’t seem proficient with OOP, leading to lots of copy/paste instead of reuse. It also doesn’t attempt to clean up tech debt or simplify code, and complex code with tech debt makes subsequent development much more difficult.

I believe that with AI, we can actually “make software development fun again.” People can focus more on the interesting parts of software development while delegating the actual code writing to AI. The result is a 5x-10x productivity gain — work that previously took 6 months can now be completed in 1 month, which feels amazing.

The future model is about how humans can better leverage AI to complete software building work. A software developer’s core competence should be acting as a CTO or team lead, guiding AI to write specific code — much like a star chef at a restaurant doesn’t wash and chop vegetables themselves. Those tasks are delegated to assistants, while the chef focuses on the core work.