AI entered our lives so fast. Keeping up with its pace takes real effort. What started as a simple tool for asking questions has completely shifted how we do software engineering. This is how I currently see where we are and where things are heading.
Syntax used to be the hard part. Today, it's usually the easiest problem in the room. As @karpathy said, the new programming language is English. At first, it sounded like a fancy quote. Now, it feels real.
Writing code alone is no longer enough. The product and the customer are now at the center. You need to clearly understand customer needs and what's actually worth improving. Even in large-scale companies where you're part of a specialized team, you still need this product-oriented mindset.
By "product-oriented mindset," I'm talking about communication, understanding the user, and getting more involved in the business side. If you don't understand why you're building something, the code doesn't really matter.
Two things matter going forward:
Working with AI Agents
With models like Opus 5 coming and sub-agents working in parallel, our job is becoming more about orchestration. We need to get better at managing token usage and designing systems where these agents can actually be effective. This feels like a different kind of engineering.
Becoming a "Product Engineer"
We should stop labeling ourselves strictly as frontend, backend, or full-stack. Product Engineer is the role that survives. This means learning the business, talking to the product teams more, and really understanding the customer. In a startup, this even extends to marketing or other areas. It's about understanding what actually matters.
My approach (and what I'm doing myself) is to follow the major shifts in AI without trying to adopt everything. A big part of this is clear communication (not just with agents), but around product intent and what actually matters to users. I'm focusing on better prompting and using LLM tools more effectively (with MCP, Skills, and Rules). Instead of chasing every new hype, I'm investing in improving my daily workflow. If something truly ground-breaking appears, we can come back to it. For now, the real priority is learning how to work with agents.