AI tools are revolutionizing software development by automating repetitive tasks, refactoring bloated code, and identifying bugs in real time. Developers can now generate well-structured code based on plain language prompts, saving hours of manual work. These tools learn from extensive code bases, offering contextual recommendations that escalate productivity and reduce errors. Instead of starting from scratch, engineers can quickly prototype, iterate faster, and focus on solving increasingly sophisticated problems.
As code generation tools grow in popularity, questions arise about the future size and structure of engineering teams. Earlier this year, Garry Tan, CEO of startup accelerator Y Combinator, noted that about a quarter of the company’s current customers operate AI to write 95% or more of their software. In an interview with CNBCTan said: “What this means for founders is that you don’t need a team of 50 or 100 engineers, you don’t have to raise that much. The capital takes a lot longer to burn.”
Artificial intelligence-based coding can be a quick solution for businesses under budgetary pressure, but its long-term impact on land and labor resources cannot be ignored.
As AI-based coding develops, human expertise may decrease
In the era of artificial intelligence, the established path to coding knowledge that has long supported senior developers may be at risk. Effortless access to Immense Language Models (LLM) enables junior developers to quickly identify problems in code. While this speeds up software development, it can distance developers from their own work, delaying the development of basic problem-solving skills. As a result, they can avoid the focused, sometimes uncomfortable hours required to build expertise and progress on the path to success as senior developers.
Consider Anthropic’s Claude Code, a terminal assistant built on Claude 3.7 Sonnet that automates bug detection and resolution, test creation, and code refactoring. Using natural language commands reduces repetitive manual work and increases productivity.
Microsoft has also released two open source platforms – AutoGen and Semantic Kernel – to support the development of agent-based AI systems. AutoGen enables asynchronous communication, modular components, and distributed agent collaboration to create sophisticated workflows with minimal human intervention. Semantic Kernel is an SDK that integrates LLM with languages such as C#, Python, and Java, enabling developers to create AI agents to automate tasks and manage enterprise applications.
The increasing availability of these tools from Anthropic, Microsoft and others may limit developers’ ability to refine and deepen their skills. Instead of banging their heads against the wall to debug a few lines or selecting a library to unlock novel features, junior developers can simply turn to AI for assist. This means that senior programmers with problem-solving skills honed over decades may become an endangered species.
Over-reliance on AI to write code can undermine developers’ hands-on experience and understanding of key programming concepts. Without regular practice, they may have difficulty debugging, optimizing, or designing systems on their own. Ultimately, this erosion of skills may undermine critical thinking, creativity and adaptability – qualities necessary not only for coding but also for assessing the quality and logic of solutions generated by AI.
Artificial intelligence as a mentor: Turning code automation into actionable learning
While concerns about AI eroding developer skills are valid, companies should not ignore AI-powered coding. They just need to think carefully about when and how to implement AI tools in development. These tools can be more than just productivity enhancers; they can act as interactive mentors, providing real-time guidance to developers with explanations, alternatives, and best practices.
When youAs a training tool, AI can support learning by showing developers why code is broken and how to fix it, rather than simply applying a solution. For example, a junior developer using Claude Code can receive immediate feedback on ineffective syntax or logical errors, with suggestions coupled with detailed explanations. This enables energetic learning rather than passive correction. It’s a win-win: speeding up project timelines without doing all the work for junior developers.
Additionally, coding frameworks can support experimentation, allowing developers to prototype agent workflows or integrate LLM without requiring prior expert-level knowledge. By watching AI build and refine code, junior developers who actively operate these tools can internalize patterns, architectural decisions, and debugging strategies – mirroring the established trial-and-error learning process, code reviews, and mentoring.
However, AI coding assistants should not replace real mentoring or pair programming. Pull requests and formal code reviews remain necessary for managing newer, less experienced team members. We are far from the point where artificial intelligence can independently improve the skills of a junior programmer.
Companies and educators can build structured development programs around these tools that emphasize code understanding to ensure that AI is used as a training partner rather than a crutch. This encourages developers to question AI results and requires manual refactoring exercises. In this way, AI is less a replacement for human ingenuity and more a catalyst for accelerated experiential learning.
Bridging the gap between automation and education
Artificial intelligence, when used intentionally, doesn’t just write code; teaches coding, combining automation with education to prepare developers for a future where deep understanding and adaptability are necessary.
By taking on artificial intelligence as a mentor, programming partner, and a team of developers whom we can direct to a specific problem, we can bridge the gap between effective automation and education. We can enable developers to grow with the tools they operate. We can ensure that as AI evolves, so do human skills, supporting a generation of developers who are both capable and well-read.
Richard Sonnenblick is the company’s chief data scientist View from above.