Software engineer and developer tool creator Jake Lundberg reports a growing trend of AI-generated content flooding professional communication and codebases, drawing parallels to the noise levels observed during the early social media boom (2008–2012). The author recounts receiving an automated, machine-generated comment on a blog post that summarized the post's own points and suggested a tool he had already announced. Because the comment appeared human-like, the author felt compelled to reply to maintain professional decorum, illustrating how the effort cost of generating low-quality content shifts from the producer to the recipient, who must spend time judging its validity.

This phenomenon, where the cost of judging content remains high while the cost of producing it becomes near zero, forces the burden downstream. In the context of software development, this leads to higher volumes of code that often lack thoughtful engineering. The author notes that generated code frequently exhibits a 'surface polish'—clean formatting and logical naming—that masks a lack of genuine consideration. This deceptive quality allows poor-quality code to bypass standard manual reviews, which are designed to spot signs of hurried work.

Evidence of this trend is supported by data from GitClear, which analyzed 211 million lines of code changed between 2020 and 2024. The data shows that the share of code marked as 'copy-pasted' has steadily climbed, while code that is refactored (cleaned up and reused) has fallen. In 2024, the volume of copy-pasted code surpassed that of refactored code for the first time. The author argues that while bad social media posts disappear as they slide off a user's feed, bad code is merged into a repository where it accrues 'interest' over time. This technical debt compounds as developers build new features on top of poor-quality code, creating a long-term maintenance burden that persists long after the original author has moved on from the project.

Software engineer and developer tool creator Jake Lundberg reports a growing trend of AI-generated content flooding professional communication and codebases, drawing parallels to the noise levels observed during the early social media boom (2008–2012). The author recounts receiving an automated, machine-generated comment on a blog post that summarized the post's own points and suggested a tool he had already announced. Because the comment appeared human-like, the author felt compelled to reply to maintain professional decorum, illustrating how the effort cost of generating low-quality content shifts from the producer to the recipient, who must spend time judging its validity.

This phenomenon, where the cost of judging content remains high while the cost of producing it becomes near zero, forces the burden downstream. In the context of software development, this leads to higher volumes of code that often lack thoughtful engineering. The author notes that generated code frequently exhibits a 'surface polish'—clean formatting and logical naming—that masks a lack of genuine consideration. This deceptive quality allows poor-quality code to bypass standard manual reviews, which are designed to spot signs of hurried work.

Evidence of this trend is supported by data from GitClear, which analyzed 211 million lines of code changed between 2020 and 2024. The data shows that the share of code marked as 'copy-pasted' has steadily climbed, while code that is refactored (cleaned up and reused) has fallen. In 2024, the volume of copy-pasted code surpassed that of refactored code for the first time. The author argues that while bad social media posts disappear as they slide off a user's feed, bad code is merged into a repository where it accrues 'interest' over time. This technical debt compounds as developers build new features on top of poor-quality code, creating a long-term maintenance burden that persists long after the original author has moved on from the project.