Boris Cherny says we need to rethink programming

Computer History Museum · CHM LiveDavid C. Brockwith Boris Cherny (Anthropic)September 23, 2026Watch the video
Boris Cherny, the developer who created Claude Code at Anthropic, thinks AI is changing how we program, which skills we need, and how we should approach building new things. In a conversation at the Computer History Museum on September 23, 2026, he explains how habits formed through traditional software development can get in the way and why curiosity and good judgment are becoming more important.
Experienced developers have to think differently
Cherny says some of his own poor decisions about AI came from applying habits he picked up in traditional software development.
Traditional software is relatively predictable. Developers write code and usually know how their system will behave. AI models behave differently, so methods that work for traditional software can be less reliable.
He says developers with decades of experience sometimes make the wrong call because they are accustomed to how things used to work. Unlearning those habits can take up to six months. He has also seen recent graduates who never formed those habits come up with better solutions than he did.
Cherny says building with AI involves more experimentation. You try different approaches, observe the results, and see what works. He likens the process to social science.
Writing code matters less
Cherny expects writing code to become less important as curiosity, good judgment, and an understanding of other fields become more valuable.
He notes that large tech companies have traditionally encouraged engineers to stay within their area of expertise. He sees less value in those boundaries as AI makes it easier to work across programming, design, analysis, and business development.
He wants developers who are curious about the things they build and the people they build them for. He also values people who can make sound decisions without constant supervision.
Everyone could soon code as well as he does
Cherny makes a specific prediction. Within a year of the interview, and perhaps sooner, he thinks anyone will be able to code as well as he can.
In his view, the gap between experienced professionals and people who have never written a line of code will keep shrinking.
He compares the shift to the printing press. Before printing, fewer people could read and write, and a small group produced most written material. Printing made it possible to distribute knowledge on a much larger scale.
Cherny thinks automating programming could give more people a similar opportunity to build software of their own. He wonders what this could mean over the next ten, twenty, or fifty years.
Does AI make us less capable?
During the interview, Cherny is asked whether Claude Code might make developers less capable by taking over work they used to do themselves.
He brings up calculators. He once multiplied large numbers in his head. Now he uses the calculator on his phone. He also used to memorize his friends’ phone numbers, which he can now look up.
His point is that technology has long taken over tasks people once had to do themselves. It frees up time and attention for learning other things and working on different problems. For Cherny, this is part of an ongoing shift in how people and computers divide up the work.
Even Claude Code could become obsolete
Cherny says he looks forward to the day his team can put Claude Code aside because it has found a better way to work with AI.
He thinks models are advancing so quickly that the tools built around them could eventually hold back what the models can do.
He also sees the boundaries between programming and other kinds of work fading. The same AI agent can write code, analyze data, do research, and help with design. A user can describe the outcome they want and let an agent handle parts of the work that once required separate tools.
We still need to understand computers
Cherny still sees value in learning how programming works as AI takes on more of the work.
He compares it to biology. Most people rarely draw on detailed knowledge of cells in everyday life, yet a basic understanding of biology is useful. He sees the same value in understanding computers and programming, even for people who rarely write code themselves.
He hopes computer science will eventually be something everyone learns in school, alongside subjects like math and biology.
What to take away
Cherny argues that we need to adapt how we think and work as AI can carry out more tasks. Experience still helps, and some habits are worth revisiting. Knowing how to experiment, learn, understand other fields, and make good decisions becomes more important.
He expects programming to become accessible to many more people, including those who have never learned to code. As technology takes on more of the work, people still need to think through what they want to build and decide whether the result is right.