{"id":1256,"date":"2026-08-29T20:47:00","date_gmt":"2026-08-30T01:47:00","guid":{"rendered":"https:\/\/www.jkspeaks.com\/wordpress\/?p=1256"},"modified":"2026-09-06T14:30:40","modified_gmt":"2026-09-06T19:30:40","slug":"claude-12-things-i-learned-the-hard-way-using-claude","status":"publish","type":"post","link":"https:\/\/www.jkspeaks.com\/wordpress\/consulting\/claude-12-things-i-learned-the-hard-way-using-claude\/","title":{"rendered":"Claude: 12 things I learned the hard way using Claude"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">\u2728Purely based on my hands-on experience building and iterating on a real-world project with Claude where actual dollars are on the line.<br \/><br \/>Passing the Claude Architect exam gave me the foundational framework, but putting it into practice revealed how subtle configuration choices directly impact token consumption, cost, and efficiency.<br \/><br \/>\ud83d\udccc<strong>Key areas where small shifts made the biggest difference<\/strong>:<br \/>\u2022 Tool overhead &amp; context window bloat<br \/>\u2022 Sub-agent architecture vs. main session costs<br \/>\u2022 Cache breakage when switching models mid-stream<br \/>\u2022 Compaction thresholds &amp; persistent <a href=\"http:\/\/memory.md\/\">memory.md<\/a> logging<br \/><br \/>Swipe through for 12 practical lessons I learned. Share your experience in the comments.<\/p>\n\n\n\n<div data-wp-interactive=\"core\/file\" class=\"wp-block-file\"><object data-wp-bind--hidden=\"!state.hasPdfPreview\" hidden class=\"wp-block-file__embed\" data=\"https:\/\/www.jkspeaks.com\/wordpress\/wp-content\/uploads\/2026\/09\/12-learnings-using-claude-v1.pdf\" type=\"application\/pdf\" style=\"width:100%;height:600px\" aria-label=\"Embed of 12 learnings using claude v1.\"><\/object><a id=\"wp-block-file--media-ef8c1682-2c5d-4e96-b21f-09abd50e61be\" href=\"https:\/\/www.jkspeaks.com\/wordpress\/wp-content\/uploads\/2026\/09\/12-learnings-using-claude-v1.pdf\">12 learnings using claude v1<\/a><a href=\"https:\/\/www.jkspeaks.com\/wordpress\/wp-content\/uploads\/2026\/09\/12-learnings-using-claude-v1.pdf\" class=\"wp-block-file__button wp-element-button\" download aria-describedby=\"wp-block-file--media-ef8c1682-2c5d-4e96-b21f-09abd50e61be\">Download<\/a><\/div>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/www.linkedin.com\/search\/results\/all\/?keywords=%23claudeai&amp;origin=HASH_TAG_FROM_FEED\">#ClaudeAI<\/a> <a href=\"https:\/\/www.linkedin.com\/search\/results\/all\/?keywords=%23enterpriseai&amp;origin=HASH_TAG_FROM_FEED\">#EnterpriseAI<\/a> <a href=\"https:\/\/www.linkedin.com\/search\/results\/all\/?keywords=%23agenticai&amp;origin=HASH_TAG_FROM_FEED\">#AgenticAI<\/a> <a href=\"https:\/\/www.linkedin.com\/search\/results\/all\/?keywords=%23aiengineering&amp;origin=HASH_TAG_FROM_FEED\">#AIEngineering<\/a> <a href=\"https:\/\/www.linkedin.com\/search\/results\/all\/?keywords=%23generativeai&amp;origin=HASH_TAG_FROM_FEED\">#GenerativeAI<\/a> <a href=\"https:\/\/www.linkedin.com\/search\/results\/all\/?keywords=%23llmops&amp;origin=HASH_TAG_FROM_FEED\">#LLMOps<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u2728Purely based on my hands-on experience building and iterating on a real-world project with Claude where actual dollars are on the line. Passing the Claude Architect exam gave me the foundational framework, but putting it into practice revealed how subtle configuration choices directly impact token consumption, cost, and efficiency. \ud83d\udcccKey areas where small shifts made&#8230;<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"_monsterinsights_skip_tracking":false,"_kad_post_transparent":"","_kad_post_title":"","_kad_post_layout":"","_kad_post_sidebar_id":"","_kad_post_content_style":"","_kad_post_vertical_padding":"","_kad_post_feature":"","_kad_post_feature_position":"","_kad_post_header":false,"_kad_post_footer":false,"_kad_post_classname":"","footnotes":"","_links_to":"","_links_to_target":""},"categories":[4],"tags":[123,179,177],"class_list":["post-1256","post","type-post","status-publish","format-standard","hentry","category-consulting","tag-ai","tag-claude","tag-experience"],"acf":{"phase":"3","cluster":"Technical","topics":["AI Architecture","AI Engineering","Claude","Ops","Orchestration"]},"_links":{"self":[{"href":"https:\/\/www.jkspeaks.com\/wordpress\/wp-json\/wp\/v2\/posts\/1256","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.jkspeaks.com\/wordpress\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.jkspeaks.com\/wordpress\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.jkspeaks.com\/wordpress\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.jkspeaks.com\/wordpress\/wp-json\/wp\/v2\/comments?post=1256"}],"version-history":[{"count":1,"href":"https:\/\/www.jkspeaks.com\/wordpress\/wp-json\/wp\/v2\/posts\/1256\/revisions"}],"predecessor-version":[{"id":1258,"href":"https:\/\/www.jkspeaks.com\/wordpress\/wp-json\/wp\/v2\/posts\/1256\/revisions\/1258"}],"wp:attachment":[{"href":"https:\/\/www.jkspeaks.com\/wordpress\/wp-json\/wp\/v2\/media?parent=1256"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.jkspeaks.com\/wordpress\/wp-json\/wp\/v2\/categories?post=1256"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.jkspeaks.com\/wordpress\/wp-json\/wp\/v2\/tags?post=1256"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}