{"id":12841,"date":"2026-07-30T20:35:17","date_gmt":"2026-07-30T12:35:17","guid":{"rendered":"http:\/\/longzhuplatform.com\/?p=12841"},"modified":"2026-07-30T20:35:17","modified_gmt":"2026-07-30T12:35:17","slug":"head-of-anthropics-claude-code-says-prompt-engineering-not-that-important-via-sejournal-martinibuster","status":"publish","type":"post","link":"http:\/\/longzhuplatform.com\/?p=12841","title":{"rendered":"Head Of Anthropic\u2019s Claude Code Says Prompt Engineering Not That Important via @sejournal, @martinibuster"},"content":{"rendered":"<p><\/p> <div id=\"narrow-cont\"> <p>Y Combinator\u2019s Diana Hu recently interviewed Boris Cherny, the creator and head of Claude Code at Anthropic. Cherny said that Claude Code is more capable than is commonly understood, saying that prompt engineering is largely not important. Given that Claude doesn\u2019t require \u201coverthinking,\u201d he suggested Claude Code users would be better off unfollowing AI influencers on Twitter and LinkedIn.<\/p> <h2>Prompt Engineering Is Less Relevant Today<\/h2> <p>One point that Cherny was excited about is that Claude Code is highly capable and does not require meticulous prompts. He gave the example of an experiment that went viral internally in which an engineer gave Claude access to OpenCV (Open Source Computer Vision Library), an open-source library of several thousand real-time vision algorithms for processing images and machine learning. Then it was asked to draw images, and it did, a capability that was completely unexpected.<\/p> <p><iframe class=\"sej-iframe-auto-height\" id=\"in-content-iframe\" scrolling=\"no\" src=\"https:\/\/www.searchenginejournal.com\/wp-json\/sscats\/v2\/tk\/Middle_Post_Text\"><\/iframe><\/p> <p>The point of the anecdote was to drive home the message that Claude Code has capabilities that far outstrip the need for \u201cprompt engineering.\u201d Diana Hu asked about model elicitation, which is the process of unlocking hidden capabilities in AI models.<\/p> <p><em>Diana Hu asked:<\/em><\/p> <blockquote> <p>\u201cAnd the big area of research for this is basically model elicitation, right? Becoming really good at figuring out all these capabilities and asking the model to do the right thing, right?<\/p> <p>How do people get better at that? And effectively, how do people get better at prompt engineering? Do people still need to do a lot of prompt engineering or is that changing as well?<\/p> <p>Tell us about where this is going.\u201d<\/p> <\/blockquote> <p>One of the things Cherny kept repeating is that it\u2019s useful to give Claude Code tasks that are hard to do and that go beyond your expectations of what Claude is capable of, which was the point of the OpenCV anecdote.<\/p> <p><em>Cherny answered:<\/em><\/p> <blockquote> <p>\u201cYeah, I remember like a year ago, one of the most popular job openings was prompt engineer. And then it kind of changed and then I think it became like context engineer. So there\u2019s these kind of waves of it, I think these will kind of like come and go.<\/p> <p>I think the skill nowadays is less about prompt engineering and more about figuring out how do you give Claude a hard task that seems a little bit too hard? And then how do you make it possible for Claude to verify its work along the way?<\/p> <p>And the verification, I think, is probably the single most important thing that people do not get right, largely.\u201d<\/p> <\/blockquote> <p>At this point, Cherny described giving Claude Tag (Claude working as an agent within Slack) access to a Mac virtual machine and then gave it a complex programming task. He asked it to rewrite the Claude desktop app, transforming it from the open-source Electron framework to the native Apple programming language Swift. Cherny used a simple, direct prompt that did not overthink the instructions or constraints.<\/p> <p><em>Here\u2019s his description of the task and the prompt:<\/em><\/p> <blockquote> <p>\u201cAnd then I was like, OK, now what I want you to do is I want you to rewrite the Electron app in Swift. I want you to run the Electron app in the Mac virtual machine, screenshot it, and then look pixel by pixel, compare it to the Swift version.<\/p> <p>Don\u2019t stop until you\u2019re done.<\/p> <p>\u2026So this is really one of those examples where the model can do it today, you just have to let it do it, and you don\u2019t need the fancy stuff.<\/p> <p>You don\u2019t need \/goal, you don\u2019t need \/loop. These help. But really all you need is give the model the task, give it a way to verify the output of its work so it doesn\u2019t get stuck, and it\u2019ll just go.<\/p> <p>\u2026And actually, in this case, Claude also decided to live blog it. So what it did is it created a Slack channel internally, and it started just posting screenshots every few minutes of its progress.\u201d<\/p> <\/blockquote> <h2>How Claude Code Users Can Become More Expert<\/h2> <p>The interviewer, Diana Hu, expressed astonishment that a complex programming task could be accomplished with a simple prompt. She next asked how Anthropic Claude Code users can become more expert with Claude Code. Cherny\u2019s answer said that prompting Claude Code has to be done \u201cempirically,\u201d which means through experience and observing how things work out.<\/p> <p><em>She asked:<\/em><\/p> <blockquote> <p>\u201cWhat is separating the people here that can become the top 1% Claude users? How can people learn to use Claude like Boris?\u201d<\/p> <\/blockquote> <p><em>Cherny replied:<\/em><\/p> <blockquote> <p>\u201cMaybe like, don\u2019t listen to the LinkedIn influencers. Don\u2019t listen to, don\u2019t read Twitter.<\/p> <p>This is the thing about the model is I think everyone\u2019s looking for like the one weird trick to do it. That doesn\u2019t exist, there\u2019s nothing like that.<\/p> <p>The way the model works is you have to approach it empirically. You have to give it a task that\u2019s too hard.<\/p> <p>You have to give it the tools to verify the work, like you would yourself, like you would if you were doing the task.<\/p> <p>You have to see where it struggles, and then you have to fix that, either with better prompting or with a skill, or if the model is missing context, give it an MCP so it can pull in the context that it needs. That\u2019s kind of it.\u201d<\/p> <\/blockquote> <h2>Over-Specifying Is Self-Defeating<\/h2> <p>The interviewer, Hu, remarked that it sounded simple. Cherny replied that he thinks that \u201cpeople tend to overthink it a little bit.\u201d His point was that over-specifying is not necessary with today\u2019s Claude Code.<\/p> <p><em>He continued:<\/em><\/p> <blockquote> <p>\u201cI think people tend to over-engineer. Because I think, in a lot of ways, when we built systems in the past, that\u2019s the way you had to do it.<\/p> <p>So when I look at engineers that have been coding for a long time, for years or for decades, this is a really, really common failure mode, is trying to over-specify, and it\u2019s trying to be overly specific, and get the model to do the task exactly the way that you would have done it.<\/p> <p>And that\u2019s just not the way the model works.<\/p> <p>But I think a lot of people are unlearning this, and it\u2019s a journey to unlearn it. It\u2019s a journey to kind of figure out how do you treat this thing like you would a coworker.<\/p> <p>I think that\u2019s the level of intelligence that it\u2019s at now.\u201d<\/p> <\/blockquote> <h2>Takeaways<\/h2> <ul> <li>Prompt engineering matters less than giving Claude clear instructions and ways to verify its work.<\/li> <li>Claude has advanced to the point where many \u201cprompt engineering\u201d tricks are no longer necessary.<\/li> <li>Claude Code is capable of more than many users realize.<\/li> <li>Challenge Claude Code with harder tasks.<\/li> <li>Giving things a try and seeing how they work out can be useful.<\/li> <li>Don\u2019t overthink prompts.<\/li> <\/ul> <h3>Watch Boris Cherny\u2019s Interview:<\/h3> <p class=\"vcont\"><iframe loading=\"lazy\" title=\"Boris Cherny: We Cut 80% of Claude Code\u2019s Prompt\" width=\"640\" height=\"360\" src=\"https:\/\/www.youtube.com\/embed\/qyPCVqFUyDo?feature=oembed\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share\" referrerpolicy=\"strict-origin-when-cross-origin\" allowfullscreen><\/iframe><\/p> <p><em>Featured Image by Shutterstock\/RixAiArt<\/em><\/p> <\/div> <p>Generative AI,News#Anthropics #Claude #Code #Prompt #Engineering #Important #sejournal #martinibuster1785414917<\/p> ","protected":false},"excerpt":{"rendered":"<p>Y Combinator\u2019s Diana Hu recently interviewed Boris Cherny, the creator and head of Claude Code at Anthropic. Cherny said that Claude Code is more capable than is commonly understood, saying that prompt engineering is largely not important. Given that Claude doesn\u2019t require \u201coverthinking,\u201d he suggested Claude Code users would be better off unfollowing AI influencers [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":12842,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[16],"tags":[12287,6258,257,259,305,415,10595,80],"class_list":["post-12841","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-accessibility","tag-anthropics","tag-claude","tag-code","tag-engineering","tag-important","tag-martinibuster","tag-prompt","tag-sejournal"],"acf":[],"_links":{"self":[{"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=\/wp\/v2\/posts\/12841","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=12841"}],"version-history":[{"count":0,"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=\/wp\/v2\/posts\/12841\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=\/wp\/v2\/media\/12842"}],"wp:attachment":[{"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=12841"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=12841"},{"taxonomy":"post_tag","embeddable":true,"href":"http:\/\/longzhuplatform.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=12841"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}