Anthropic’s Boris Cherny Says AI Coding Bugs Are Getting Harder: ‘Less Off-by-Ones,’ More System Design Problems
Boris Cherny, the engineer behind Anthropic‘s Claude Code, says AI coding tools are getting better at handling routine programming errors, but developers are increasingly confronting harder problems involving system design, UI usability and broader context.
AI Coding Bugs Becoming More Complex
On Tuesday, Cherny said in a post on X that large language models still produce software bugs, but the nature of those errors has changed as AI coding capabilities have improved.
"LLMs still produce bugs, but those bugs are different than what they used to be," Cherny said.
He added that problems are now "less off-by-ones and more about system design, UI usability, missing broader context."
He said some coding tasks have effectively been solved by AI, but cautioned that "not all" software engineering problems have been addressed.
Cherny also highlighted adversarial code review as a way to identify problems that may escape conventional AI-generated coding workflows.
"Adversarial code review has been an incredibly powerful tool to catch many of these kinds of bugs," he said.
As an example, Cherny suggested prompting an AI coding system to "adversarial test every edge case in an iOS simulator."
He also pointed to Claude’s built-in /code-review capability and its different review levels.
AI Coding’s New Era
AI’s role in software development had expanded rapidly, with developers increasingly relying on AI agents for coding, creative projects and complex workflows.
Earlier, AI researcher Andrej Karpathy said Anthropic’s Claude Opus had generated a procedural 3D world inspired by The Lord of the Rings, producing about 5,500 lines of code in two hours.
He said such capabilities could make highly customized projects feasible that previously required significant human effort.
At Uber, CTO Praveen Neppalli said the number of employees using frontier AI tools had more than quadrupled, while the company’s cost per token had declined.
He said the industry was moving beyond "tokenmaxxing" toward more efficient AI usage.
OpenAI President Greg Brockman said AI coding tools had gone from writing 20% to 80% of developer code in one month.
Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors.
Photo courtesy: Thrive Studios ID on Shutterstock.com