Top News

The post may seem confusing at first, but it looks like he is indicating a broader shift underway in the technology industry. As AI tools are becoming more powerful, professionals around the world are now spending more time telling machines how to work, reviewing their output and correcting it. In terms of programming, this means developers are now spending less and less time writing lines of code and instead using AI tools to test and improve software.

However, the challenge is that AI tools are unlike traditional software, as they can be unpredictable at times and confidently make mistakes. Another problem is that these tools provide different answers each time, making it difficult to understand them fully.

While these tools promise significant gains in productivity, many workers, including Karpathy himself, say they are struggling to keep up. Professionals are being forced to learn new ways of working, as AI increasingly acts as a collaborator rather than a direct replacement for humans. Karpathy argues that those who adapt quickly to this shift are likely to see sharp increases in productivity, while those who fail to do so risk being left behind.

Karpathy also claims that he would be 10 times more productive if he could “properly string together what has become available over the last ~year.” But the problem is that AI tools don’t have any manuals, and people have to figure them out on the fly.