Boris Cherny
A software engineering observer who has noted the phenomenon of blurring role boundaries for engineers in the AI era, with work reorganizing around outcomes rather than traditional job definitions.
Core Facts
Timeline (last 90 days)
Boris said that with the new model and the classifier Anthropic runs on all traffic, prompt injection is hard to make work.
Boris Cherny said that in a single evening he has thousands of Agents running, Claude maintains its own codebase and runs 20 to 30 routines per day, each described with just one sentence.
The most frequent user mistake Boris has observed is over-specification (telling the AI how to think); the correct approach should shift to defining output specifications (clearly defining success criteria).
Boris Cherny, the person in charge of Claude Code, recommends deleting CLAUDE.md and points out that the new model runs more efficiently with fewer instructions.
Boris Cherny, creator of Anthropic Claude Code, said he writes 90% of his code in Slack rather than in an IDE.
Boris Cherny said on Hacker News that plan mode is no longer useful.
ClaudeCode lead Boris Cherny stated that his work now involves writing systematic looping logic rather than individual prompts
Boris's advice is aimed at startup engineers with leisure time for experimentation who work in highly technical fields, and is of limited direct use to non-technical users.
Boris Cherny, the creator of Claude Code, advises users to periodically delete the CLAUDE.md file, skills, and hooks to see what the model can do on its own.
Boris advises adding deleted instructions back only when the model is seen repeatedly getting stuck on the same thing.
6 more timeline events
All Facts (20)
The head of Claude Code is Boris Cherny.
80%UnverifiedBoris Cherny曾在Meta工作
90%UnverifiedAnthropic的Claude Code贡献者Boris Cherny宣称'软件工程已死'
75%UnverifiedBoris Cherny建议的AI编程工作流核心是构建反馈闭环：提前定义验收标准、让AI自己运行测试、发现问题自动修复、循环迭代直到通过
60%UnverifiedBoris Cherny was previously known to the developer community for his book Programming TypeScript.
60%UnverifiedBoris said that with the new model and the classifier Anthropic runs on all traffic, prompt injection is hard to make work.
50%UnverifiedBoris Cherny, the person in charge of Claude Code, recommends deleting CLAUDE.md and points out that the new model runs more efficiently with fewer instructions.
50%UnverifiedThe most frequent user mistake Boris has observed is over-specification (telling the AI how to think); the correct approach should shift to defining output specifications (clearly defining success criteria).
50%UnverifiedBoris Cherny said that in a single evening he has thousands of Agents running, Claude maintains its own codebase and runs 20 to 30 routines per day, each described with just one sentence.
50%UnverifiedBoris Cherny, creator of Anthropic Claude Code, said he writes 90% of his code in Slack rather than in an IDE.
50%UnverifiedBoris Cherny said on Hacker News that plan mode is no longer useful.
50%UnverifiedClaudeCode lead Boris Cherny stated that his work now involves writing systematic looping logic rather than individual prompts
50%UnverifiedBoris's advice is aimed at startup engineers with leisure time for experimentation who work in highly technical fields, and is of limited direct use to non-technical users.
50%UnverifiedBoris Cherny, the creator of Claude Code, advises users to periodically delete the CLAUDE.md file, skills, and hooks to see what the model can do on its own.
50%UnverifiedBoris advises adding deleted instructions back only when the model is seen repeatedly getting stuck on the same thing.
50%UnverifiedThe core principle of context engineering optimization is to retain all information specific to the user's situation and remove restrictive rules.
50%UnverifiedThe concept of Loop Engineering was proposed by OpenAI's Boris Cherny and others, as well as the core developers of Claude Code.
50%UnverifiedAnthropic engineer Boris Cherny argues that code written by Claude and deployed to production should be subject to a higher bar of scrutiny than code written by humans.
50%UnverifiedBoris Cherny warns that without these guardrail mechanisms, the end result will likely be a mess that is difficult to maintain over the long term.
50%UnverifiedAnthropic's Boris Cherny and his team published the AI-Native SDLC Handbook.
50%