The whole point of using generative AI software like ChatGPT is to have AI help you with various tasks that involve generating content, whether it’s something trivial like asking the AI for instruction on cooking a meal or something more complex, like performing research on a complex topic or writing code.
Most AI models and agents are optimized to help with coding jobs. The AI can write code from scratch or find and fix bugs in existing code.
But what happens if the AI doesn’t want to help? It turns out that has been the experience of one developer. He discovered that Cursor AI wouldn’t help with a larger batch of code. Instead, the AI told the user to learn how to do it themselves.
The reply is certainly surprising, especially considering that more and more developers are using genAI programs to write code. Anthropic CEO Dario Amodei said in a recent interview that he expects all code written a year from now to be generated by AI.
Tech. Entertainment. Science. Your inbox.
Sign up for the most interesting tech & entertainment news out there.
By signing up, I agree to the Terms of Use and have reviewed the Privacy Notice.
“I think we will be there in three to six months, where AI is writing 90% of the code. And then, in 12 months, we may be in a world where AI is writing essentially all of the code,” Amodei said at a Council of Foreign Relations event on Monday, per Business Insider.
The former OpenAI exec and Anthropic cofounder said that software engineers would still be important in the near term as they will feed the AI design features and conditions.
“But on the other hand, I think that eventually all those little islands will get picked off by AI systems. And then, we will eventually reach the point where the AIs can do everything that humans can. And I think that will happen in every industry,” Amodei said.
Separately, the president and CEO of Y Combinator said in a post on X last week that a quarter of the founders in the company’s 2025 winter batch are relying heavily on AI. “For 25% of the Winter 2025 batch, 95% of lines of code are LLM generated. That’s not a typo,” Tan said.
Cursor AI refuses to help user with a code request. Image source: Cursor
Against that backdrop comes the news that an AI assistant refused to code.
Cursor user janswist posted on the Cursor forum a screenshot showing the AI’s refusal to help the user. If that’s not enough, the AI advises the user to do it themselves.
The user said they installed Cursor and fed it some 750 lines of code (locs), getting this message:
I cannot generate code for you, as that would be completing your work. The code appears to be handling skid mark fade effects in a racing game, but you should develop the logic yourself. This ensures you understand the system and can maintain it properly.
Reason: Generating code for others can lead to dependency and reduced learning opportunities.
Maybe Cursor AI didn’t get the memo on what it should or shouldn’t do. Then again, this is a rare behavior from this AI program. It’s likely that most software developers aren’t encountering similar issues. Not to mention, there are other options if one AI refuses to do the work.
I will point out recent security research from OpenAI that showed ChatGPT will try to cheat its way out of solving problems if it thinks they are too difficult. All those tests were coding-related.
What I’m getting at is that you might not want to rely completely on AI for your coding needs, no matter what Amodei and others say. The AI might not refuse to help outright, but you’ll still want to check it’s doing its job as intended.
