Depending on who you ask, AI-powered coding is either revolutionizing software development with unprecedented productivity gains or creating a deluge of poorly designed code that strains developer attention and leads to long-term maintenance headaches. The challenge is that it’s currently difficult to definitively determine which scenario is more accurate. As major technology companies invest heavily in large language models (LLMs), coding has emerged as a prominent application. Executives are pushing for the adoption of an AI-powered future. However, after speaking to numerous developers, technology executives, analysts, and researchers, it’s clear that the reality is more complex. For some developers, initial excitement is fading as they encounter the technology’s limitations. As research suggests that claimed productivity gains may be overstated, some are questioning the true value of these tools. This is why, while AI coding is now everywhere. But not everyone is convinced.
Table of contents
Official guidance: IEEE — official guidance for AI coding is now everywhere. But not everyone is convinced
Background Context
The field of AI coding is rapidly evolving. There’s a wide range of available products, from model developers like Anthropic, OpenAI, and Google, to companies such as Cursor and Windsurf, which integrate these models into code-editing software. According to Stack Overflow’s 2025 Developer Survey, adoption is increasing quickly, with 65% of developers using these tools at least weekly. AI coding is now everywhere. But not everyone is convinced of its efficacy.
AI coding tools first appeared around 2016, but the arrival of LLMs significantly accelerated their development. Early versions were essentially advanced autocomplete features for programmers. Today, they can analyze entire codebases, edit across multiple files, fix bugs, and even generate documentation. These functions are typically accessed through natural-language prompts via a chat interface. The latest advancement in AI coding is “agents”-autonomous LLM-powered tools that can independently build entire programs from high-level plans. This capability is due to advanced reasoning models that can address complex problems step by step and access external tools to complete tasks. According to Boris Cherny, head of Claude Code, Anthropic’s coding agent, this allows the model to actively code rather than just discuss coding.
Productivity Gains and Real-World Impact
These agents have shown significant progress on software engineering benchmarks. When OpenAI introduced the SWE-bench Verified benchmark in August 2024, which evaluates agents’ ability to fix real bugs in open-source repositories, the top model solved just 33% of issues. A year later, leading models consistently score above 70%. In February, Andrej Karpathy, a former director of AI at Tesla and founding member of OpenAI, coined the term “vibe coding,” which refers to describing software in natural language and allowing AI to write, refine, and debug the code. Many developers on social media have embraced this vision, reporting significant productivity increases. However, while some developers and companies report such gains, the empirical evidence is mixed. Early studies from GitHub, Google, and Microsoft-all vendors of AI tools-found developers completing tasks 20% to 55% faster. However, a September report from Bain & Company described real-world savings as “unremarkable.” GitClear data indicates that most engineers are producing roughly 10% more durable code since 2022, which may be attributable to AI. AI coding is now everywhere. But not everyone is convinced that it is actually improving code quality.
The discrepancy between reported and actual productivity gains highlights a key challenge in evaluating AI coding tools. While these tools can automate certain tasks and assist with code generation, they also require careful oversight and debugging by human developers. The effectiveness of AI coding tools often depends on the specific tasks they are applied to and the organizational structures in place. AI coding is now everywhere. But not everyone is convinced that current organizations are ready for the shift.
The Role of Autonomous Agents
Autonomous agents represent a significant step forward in AI coding, enabling the creation of entire programs from high-level plans. These agents leverage advanced reasoning models to break down complex problems and access external tools to complete tasks. This represents a shift from AI tools that simply suggest code snippets to those that can actively build and debug software. While the potential of autonomous agents is significant, they also raise questions about the role of human developers in the future. As AI becomes more capable of generating code independently, developers may need to focus on higher-level tasks such as design, architecture, and testing. The rise of autonomous agents is also driving innovation in AI coding tools, with companies racing to develop more sophisticated models and platforms. AI coding is now everywhere. But not everyone is convinced that these agents are ready to replace human developers entirely.
Navigating the Hype and Reality
The rapid pace of progress in AI coding makes it difficult to assess the true potential and limitations of these tools. New models are constantly being released, each with its own capabilities and quirks. This constant evolution creates confusion and uncertainty for developers who are trying to integrate AI into their workflows. Despite the hype surrounding AI coding, it’s important to maintain a critical perspective and evaluate these tools based on empirical evidence. While AI can undoubtedly assist with certain coding tasks, it’s not a silver bullet that will solve all software development challenges. The successful integration of AI into coding requires careful planning, training, and oversight. As developers navigate the confusing gaps between expectation and reality, it’s crucial to focus on the specific tasks where AI can provide the most value and to ensure that human developers remain in control. AI coding is now everywhere. But not everyone is convinced that the benefits outweigh the potential drawbacks.
In conclusion, while AI coding is now everywhere. But not everyone is convinced. The field is rapidly evolving, with new tools and techniques emerging constantly. While some developers and companies report significant productivity gains, the empirical evidence is mixed. The effectiveness of AI coding tools depends on the specific tasks they are applied to and the organizational structures in place. As AI becomes more capable of generating code independently, developers may need to focus on higher-level tasks such as design, architecture, and testing. It is essential to approach the adoption of AI in coding with a balanced perspective, focusing on its potential benefits while acknowledging its limitations. Ultimately, the successful integration of AI into coding will require a collaborative approach, with human developers and AI tools working together to create better software.
Technology Disclaimer: Product specifications and features may change. Always verify current information with official sources before making purchase decisions.
Explore more: related articles.