AI Code Reviews: Why They're Draining Our Attention (and Us)
·2 min read·Intermediate
“
Imagine working harder, yet seeing no real quality boost. That's the AI code review paradox, leaving us utterly drained. The real culprit isn't AI itself, but our dwindling attention.
In 30 seconds
01AI in code reviews boosted work volume, but failed to improve quality.
02
→
💡
What this means for you
For us mere mortals, it means even AI, if misused, can create more stress than benefit. The quality of our attention is far more crucial than the sheer volume of information we process.
Making talking-head videos, where you're the star explaining something, can be a monumental pain. Now imagine an AI doing most of the heavy lifting, right there in your browser.
·1 min·3·Beginner
Bots now handle reviews and replies, creating an endless attention loop.
03The core issue is "Return on Attention," draining developer energy invisibly.
0101
What's the problem with AI code reviews?
AI-generated code reviews are actually increasing team workload without a corresponding jump in software quality. It seems we're just giving ourselves more stuff to read.
Pull requests used to be about improving code. Now, with AI churning them out non-stop, their volume has skyrocketed. The catch? The initial ticket quality hasn't improved at all, leaving teams swamped with mediocre changes.
Large Language Models (LLMs) have filled the gap between quantity and quality. We're now seeing bots reviewing code, other bots replying, and sometimes even bots "arguing" with each other. It's an automation circus where humans are just tired spectators. In 2023, the adoption of AI tools for code reviews led to an average 30% increase in the number of pull requests managed by development teams.
0202
What is "Return on Attention"?
The real issue isn't AI reviewing code, but our "Return on Attention." We're pouring massive mental energy into processes that yield diminishing value. It's like investing a dollar and getting ten cents of useful attention back.
📬 Enjoying this article?
Get the best AI news every week, straight to your inbox.
One CEO hit the nail on the head: it's not a "code review" problem per se. It's a broader issue concerning how we allocate our finite attention span. If bots generate noise, our attention inevitably gets scattered.
Imagine sifting through dozens of AI-generated comments, some sensible, others less so. Your mind quickly fatigues. Before, you'd read targeted human reviews; now you're filtering a river of automated text. Is it worth it? A Microsoft report from Q4 2023 highlighted that employees spend approximately 40% of their time on tasks considered low-value, often amplified by AI tools.
0303
How can we avoid going crazy?
We need to stop assuming more automation is always better, especially when human attention is involved. The goal should be using AI to reduce the noise, not amplify it unnecessarily.
Instead of letting AI generate endless replies and comments, we could use it to filter out less important pull requests or summarize key points. The aim is to present only the essence, not the entire squeezed juice.
Perhaps we should focus on upstream quality. If code and tickets are better written from the start, there will be less need for bots arguing with other bots. Ultimately, the human mind remains the most valuable resource. Google, starting in February 2024, began testing AI systems for code reviews that focus on comment summarization, reducing text by 25% for human reviewers.
Imagine someone taking apart your favorite Android game, piece by piece. A new GitHub project just made that a little easier, but only for the truly dedicated tech-heads.
