Coding Is Down, Engineering Is Up

22 Jan 2026

A line I heard in a podcast snagged in my brain and I’ve been thinking about it ever since.

It said “software development is on the decline, but engineering is more critical than ever”. This is a very insightful take.

We know that writing code is getting easier. Claude Code, Cursor, Replit, Windsurf or even Lovable, if you’re a beginner like me, will create the code you want.

We can see the impact of this: GitHub Copilot users report 30–55% faster task completion for routine coding. Microsoft and Google have both said 25–30% of new code is AI-generated internally. Entry-level dev hiring is down ~20–40% across the US/UK since 2022.

But in the past, ever since the web, developers have been able to focus on writing code without having to really worry about the system or environments it’s running on, beyond a point. It’s an abstraction.

Today you can no longer restrict yourself to just the layer of code. Think about the set of layers below:

• Experience design and logic
• Application code ← this is what AI Generates.
• Model layer (LLMs, embeddings, fine-tunes, agents)
• Data layer (pipelines, vector stores, provenance, drift)
• Platform layer (APIs, orchestration, MCP, observability)
• Cloud layer (compute, storage, networking, regions)
• Infrastructure layer (GPUs, energy, latency, supply chains)
• Governance layer (security, compliance, auditability)

If you’re setting out to solve a problem today, you probably have to think through many, if not all, of this list. The onus has shifted from writing software to building systems. We are moving from bricklaying, to architecture. There are many more ‘why’ questions than ‘how’ questions. You can write a particular logic, but do you know why it exists?

Then there’s constraint management. Every business has a ton of constraints - from compliance, to legacy, to fast changing business rules. Engineering is fundamentally about working through constraints.

In some ways, engineering has had to flood the barriers of software engineering and spill over into all the other layers - architecture, security, reliability, system design. The cost of poor architecture, or security, or compliance can be sky high. And more than the business logic, the focus is now on how the layers interact with each other.

This means that the key skills may include systems thinking, intent translation, risk engineering, human machine orchestration, failure modes, and adaptability.

Ultimately, in the same way that Excel didn’t kill finance or accounts. At the bottom of the pyramid, development roles will probably erode, but the top 20% will become more far more valuable.

(PS. I am neither a software developer nor a systems engineer, this is just my observation from the ringside).