The creator of Claude Code named 3 working principles that he shares with each member of his team
Boris Cherny from Anthropic has a simple list of principles for his team, and, quite predictably, Claude is central to it.
Boris Cherny from Anthropic has a simple list of principles for his team, and, quite predictably, Claude is central to it.
Boris Cherny from Anthropic has a simple list of principles for his team, and, quite predictably, Claude is central to it.
"If you have Claude, you can really automate a huge amount of work, and that's something we see time and time again," Cherny said on a recent episode of Lenny's Podcast.
Host Lenny Raczycki noted that he had heard of one of Anthropic's principles: "What could be better than doing the work yourself? Only having Claude do it."
“Cherny noted that another of his principles is to “underfund projects a little bit,” as this forces the team to really rely on AI tools like Claude.
"When you create a small shortage of resources, an interesting effect occurs: people are simply forced to 'clodify' their workflows," he said.
While it's important to support small teams, Cherny advises CTOs not to "cut costs at the very beginning."
“Start by simply giving engineers as many tokens as possible,” he noted.
Cerny refuted the idea that Anthropic would make huge profits with this approach, especially if companies only allow a few engineers to experiment.
"Let's say they create something incredible that requires a huge number of tokens, and then the costs become significant," he said. "That's when you should start optimizing, but don't do it too early."
The cost of tokens — a component of the costs companies incur to deploy AI models — is becoming a major topic of discussion in the tech industry. OpenCode creator Dax Raad recently wrote that some CFOs are experiencing “price tag shock” when they see how much the cost per engineer is increasing due to AI bills.
The final principle is about speed: “Encourage people to move faster.” This axiom seems entirely logical when you consider how just a few weeks ago Anthropic and its competitor OpenAI released major updates to their programming tools within minutes of each other. Previously, Anthropic used Claude to build Claude Cowork, an AI agent for non-technical tasks, in just 10 days.
“It was critical at the very beginning because I was working alone, so our only advantage was speed,” Cherny said. “It was the only way to release a product that could compete in this crowded market of programming tools.”
Now Cherny is turning to Claude precisely to maintain the pace.
“This principle is still key in our team,” he says. “If you want to move faster, the best way to do it is to just give Claude more tasks. It’s a great incentive to speed up.”
Recall that recently, Borys Chernyi declared his Ukrainian origin and mentioned that his family lived in Odessa before moving to the United States.

