Agentic development is fundamentally changing the role of the developer

“Three factors contributed to that. First and foremost, the models are now genuinely capable of multi-step reasoning. Two years ago, that simply wasn’t the case.  

Secondly, the development tooling has matured significantly. IDE integrations, agent frameworks and orchestration layers have advanced to the point where you can deploy these kinds of systems into production without having to build everything from scratch.  

Finally, the cost per token has fallen dramatically. What cost around $20 per million tokens in 2022 now costs less than $0.50. That means the economic threshold has been crossed, making it financially viable to deploy agents at scale. Remove any one of those three factors and agents remain largely a research topic. It is the combination of all three that has enabled developers to use this technology in practice today.” 

“Agents will no longer be responsible for individual tasks; they will take ownership of entire domains. You’ll have a dedicated agent managing your infrastructure, another monitoring your codebase for anomalies, and another writing and executing your tests. At some point, the daily stand-up may consist of little more than reviewing status reports from your agents over a cup of coffee.  

The role of the developer will shift from writing code to defining contracts, clear inputs and outputs, and quality controls, while the agents take care of implementation. I also expect to see significant consolidation. Teams that master agent-based workflows will be five to ten times more productive than those that do not, and that will fundamentally reshape how organisations structure their engineering teams.” 

“It is already changing the way I work. Complex systems that would previously have required weeks of intensive development are now being delivered within days. Entire layers of an application, data pipelines, API integrations and test suites, can be generated, modified and debugged while agents handle the heavy lifting. 

These days, I routinely run a dozen agents in parallel to implement features, resolve bugs and take care of the repetitive work that used to dominate my schedule. As a result, I now spend a significant portion of my week on self-study, exploring new ideas and venturing further down the AI rabbit hole. That is a benefit that is difficult to quantify in monetary terms. 

The biggest shift, however, is a mental one. I have stopped optimising purely for delivery. With AI support, achieving results has become almost a given. What I optimise for now is delivering those results in the most cost-efficient way possible. With every architectural decision, every tooling choice and every agent I design, I ask myself what the most streamlined path is to the desired outcome. That means fewer over-engineered solutions, tighter scope definition and a constant awareness of where AI is doing the heavy lifting. My time and budget are directed precisely where they add the most value.”

“Agentic development is fundamentally changing the profession of software development. For thirty years, the limiting factor was how quickly you could write high-quality code. Today, the limiting factor is how well you can think.  

Success now depends on how clearly you can define a problem, break it down and design a system in which agents can operate without making costly mistakes. The best developers of the future will not be the fastest typists, nor even the strongest algorithm designers. They will be the best architects and the sharpest systems thinkers. The most valuable thing a developer can do is think clearly and drink coffee while the agents handle the rest. That is a game I genuinely find exciting to play.”