August 4, 2026
Debugging the hype machine
Eight Myths on Software Engineering and GenAI
Turns out coders may type less than you think—and the comments are fighting about it
TLDR: The article says artificial intelligence won’t magically transform software work just by writing more code, because developers spend surprisingly little of their day typing. Commenters agree the hype is overblown, but they’re arguing loudly over whether faster code-writing still reshapes everything around it.
A new ACM Queue article tried to pop the hype balloon around artificial intelligence for software work, arguing that many bosses are buying into myths. Its biggest eyebrow-raiser: studies say developers spend only about 14% of their time actually writing code. The rest? Planning, reading, debugging, reviewing, meetings, and all the invisible work that makes software function. The article’s message is basically: if you judge AI by how many lines of code it spits out, you may be measuring the wrong thing.
But the real fireworks were in the comments, where readers instantly turned this into a mini civil war over what “coding time” even means. One camp said the 14% stat feels painfully real, especially after tracking their own days and discovering that typing is just the tip of the iceberg. Another camp was not buying the article’s implied ceiling at all, arguing that if AI speeds up code writing, it can also erase some of the work that comes before it. In other words: less typing might still change the whole day.
And then came the jokes. One commenter summed up the chaos with a meme-like table and the line, “Writing Code Isn’t the Bottleneck, until writing code is the bottleneck, until it’s not again.” Another snarked that the article could have had any Fibonacci number of myths, because apparently even myth-counting is now up for debate. The mood? Equal parts skepticism, nerd humor, and exhausted “please stop oversimplifying my job” energy.
Key Points
- •The article examines eight common myths about generative AI in software engineering and argues for evidence-based adoption decisions.
- •It states that developers spend a relatively small share of their time writing code, citing a 2025 Microsoft study that found coding accounted for 14% of developer time.
- •The article argues that AI code generation therefore targets only one part of software engineering work, which also includes design, meetings, planning, and code review.
- •It says lines-of-code metrics are not a valid way to measure AI impact and are not meaningfully tied to outcomes such as software quality or delivery speed.
- •The article says AI benefits vary by task, developer, and context, and that successful adoption requires organizational workflow changes, trust, learning time, and consideration of enterprise constraints.