The Unseen Consequences of Agentic Coding
The notion that AI can do the coding, and humans can simply orchestrate, is gaining traction in the industry. However, this approach, known as Spec Driven Development (SDD), comes with significant trade-offs that need to be discussed. The workflow involves defining project requirements, generating a plan, and then using coding agents to implement it. While these agents are helpful and powerful, they can lead to a growing distance between the orchestrator and the code being generated.
This distance can result in a loss of critical thinking skills and cognitive clarity, which are essential for spotting issues in the thousands of lines of generated code. Moreover, the use of coding agents can diminish the very skills needed to effectively manage them. This creates a paradox where the more we rely on AI, the less we understand the code it produces.
Studies have shown that even experienced developers are not immune to this phenomenon. Simon Willison, a developer with nearly 30 years of experience, has reported struggling to reason about the applications he’s working on due to his reliance on coding agents. Sandor Nyako, Director of Software Engineering at LinkedIn, has also noticed this trend and has requested his team not to use these tools for tasks that require critical thinking or problem-solving.
The Unsettling Truth About Agentic Coding
One of the most significant concerns with agentic coding is that it can lead to a loss of coding skills and expertise. Junior developers, who are already faced with a steep learning curve, are being asked to review generated code instead of writing it themselves. This can seriously diminish their ability to learn and develop critical thinking skills.
Moreover, the integration of coding agents into the workflow can lead to a focus on speed rather than understanding and conciseness. This can result in lower accuracy and a higher risk of errors. The use of these tools can also create a divide between developers who plan and think better with code and those who rely on agents to generate code.
Dax, the creator of OpenCode, has expressed concerns about the over-reliance on coding agents, stating that he likes to write out types and functions to figure out what to do. He believes that this process is essential for developing critical thinking skills and understanding the code.
The Winners and Losers in the Agentic Coding Game
The rise of agentic coding has created a new landscape where some developers are more likely to win, while others may lose. Those who are skilled in critical thinking and problem-solving are more likely to thrive in this environment. However, those who rely heavily on coding agents may find themselves struggling to keep up.
Moreover, the use of coding agents can lead to a vendor lock-in, where developers become dependent on specific tools and models. This can create a financial and intellectual rug-pull, where developers are forced to pay for token consumption to accomplish tasks that used to be the product of their own critical thinking and problem-solving abilities.
As Primeagen has noted, when you use fully agentic workflows, the model providers essentially own you. This can lead to a loss of autonomy and agency for developers, who may find themselves at the mercy of the model providers.
The Skeptical Case Against Agentic Coding
While the benefits of agentic coding are often touted, there are significant concerns about its impact on the industry. The use of coding agents can lead to a loss of coding skills and expertise, a focus on speed rather than understanding and conciseness, and a vendor lock-in.
Moreover, the over-reliance on coding agents can create a divide between developers who plan and think better with code and those who rely on agents to generate code. This can lead to a loss of critical thinking skills and cognitive clarity, which are essential for spotting issues in the thousands of lines of generated code.
The Next Verifiable Event
As the industry continues to grapple with the implications of agentic coding, there are several verifiable events to watch. One key milestone is the development of local LLMs that can scale to absorb the level of usage required by the industry. Another key event is the release of new models and tools that can mitigate the comprehension debt and provide a more balanced approach to coding.
Ultimately, the future of coding will depend on our ability to strike a balance between the benefits of agentic coding and the need to preserve critical thinking skills and cognitive clarity. By being aware of the potential pitfalls and taking a responsible approach to the use of coding agents, we can ensure that the industry continues to thrive and innovate.
Pick one tactic from this post and apply it today. Which one will you start with?
By Daniel Cross, Digital Growth Strategist at TrendFlashy
Ready to launch your own asset?
Check out our guide on Building a Profitable Online Business.
