AI-First Engineering
Technology innovation used to begin with blank whiteboarding, then followed by endless meetings, prototyping and tons of hours of actual coding and deployment. But not anymore.
AI changed that.
A few years ago, we talked about AI mostly as chatbots or, at best, a glorified StackOverflow. Even a few months ago, many of us were still sceptical about using ChatGPT seriously in our daily work. I know I was. But things are changing faster than I expected. The pace at which these models are improving, combined with the inception of AI agents, feels like we’re witnessing the beginning of another major shift in our industry.
Not just new tools. A new way of working.
This talk by Fiona Fung, Anthropic, explains exactly that.
What used to be the bottleneck is no longer the bottleneck today.
Code can be churned out in seconds instead of months. Rather than going through endless meetings and approval cycles, discussions can start with three or more working prototypes already on the table.
Things that used to feel like chores — writing documentation, for example — turn out to be exactly where modern AI shines. Test-Driven Development (TDD) has always been difficult, because if we’re honest, most developers don’t enjoy starting with a failing test. But now, generating those tests is almost effortless. Ideas are no longer scarce. With AI, you can explore dozens of directions before finishing your morning coffee. Even technical debt, the parts of the system nobody wants to touch, can now be analysed, documented, and refactored with AI as a partner.
The bottleneck hasn’t disappeared. It has simply moved.
Get Weng Kee Teh’s stories in your inbox
Join Medium for free to get updates from this writer.
The challenge is no longer generating ideas or writing code. It’s deciding what is worth building, choosing among countless possibilities, and making good decisions faster than ever before.
What does it mean for us?
Gone are the days when we had to dungeon crawl through StackOverflow, piecing together answers from a dozen tabs and hoping one of them worked. AI democratizes knowledge. But it does not democratize wisdom. It has never been easier to learn something new. Whether it’s figuring out how to pick a ripe avocado, understanding a complex algorithm, or orchestrating an entire IT ecosystem with a fleet of AI agents, the answer is often just a prompt away.
The barrier to knowledge is falling rapidly. The question is no longer Can I do this? But should I do this? And perhaps more importantly: What problem is actually worth solving?
Ways of Working
There will be a shift in the way we work.
New roles are already starting to appear, such as the Forward Deployed Engineer. At the same time, it is only a matter of time before some traditional roles become obsolete. It is inevitable.
We will likely spend less time writing code line by line and more time experimenting, innovating, and failing fast. More time will also go into improving systems and refining processes rather than just building features. We need to retrain ourselves to think more like architects. At the same time, governance and policies will need to evolve. Strong guardrails are required to manage and control the power of increasingly capable AI models.
The way we work is not just changing, it is being redefined.
Future
It feels less like a single skill will define success, and more like a combination of different capabilities working together. A good understanding of AI will matter, of course. So will data science fundamentals and solid architecture thinking. But beyond the technical layer, I increasingly see leadership and strategic thinking becoming just as important. The real value comes from the ability to connect the dots between technology, people and business outcomes.
That space in between is where the most interesting problems live, and probably where the most meaningful impact will be made.
