@DanielSmidstrup Claude for figuring out what I even want to build. Codex when I know exactly what needs to happen. came from WordPress, now vibe coding in React, and that division has been pretty consistent across the stack.
@MelkeyDev makes sense from a frontend angle. Claude reads intent better when the UI has real constraints. on the React side I've noticed the same thing. it handles component architecture more gracefully than Codex when the scope grows.
@DanKulkov the funniest part is when you go to check the bill fully expecting pain and it's literally $6. rewired my brain about what "running AI in production" actually costs π
@mitsuhiko the WordPress plugin ecosystem sits right in this tension. core team patches like Debian. but the plugin repo is pure npm: build a free plugin on a weekend, 300k sites depend on it, then wonder why support requests go unanswered.
@marclou as someone who spent years on WordPress sites, this is painfully real. built a perfect API in 2 shots. then spent 45 mins arguing why 16px padding isn't 'spacious enough'.
@ashoKumar89 User enumeration is the obvious one. But even with a generic response, no rate limiting here means someone can spam thousands of reset requests and DoS your email service.π
@coder_blvck 15 years of writing every line by hand. Now I vibe code. My output tripled. If that makes me 'not a software developer' I'll happily take the L π
@alexalbert__ The 80% stat feels abstract until you realize: researchers who stopped hand-writing code didn't lose something. They moved up a layer. Same shift happening at indie builder scale too.
@iuditg the $1500 cap isn't about model access, it's about spend governance. 10,000 employees with uncapped API access and no ceiling means every team maxes it monthly with zero accountability. the cap forces actual ROI thinking before provisioning.
@hii_mohit Directory sites hit this constantly. Google sees the same template 150x and only indexes what feels unique. Add something genuinely distinct to each unindexed profile first, then submit via GSC. Otherwise you're just queuing them to get rejected again.
@ClaudeDevs This + shell scripting opens up serious automation pipelines. Already been wiring Claude API into WordPress via elementor-mcp (https://t.co/6IEufMvYnU). A proper CLI makes that dev loop way cleaner π₯
@antirez The not-delivering-is-better point is key. Shipping a worse model and marketing it as better damages trust more than a delayed release. Hope this is a one-off.
@Its_Nova1012 Everyone picks A in theory... yet every codebase I've inherited looks like B. REST conventions are the first thing that gets abandoned when a deadline hits.
@petergyang it's not just those two. .cursorrules, .windsurfrules, agents.md, CLAUDE.md. project roots are turning into AI config soup. been maintaining a single WP skills file I can import into all of them: https://t.co/AhoQ9l0RVN
@ShamashAran wait until AI coding assistants normalize it even harder. half my Claude/Cursor responses end in curl|bash now. the pattern was annoying before, it's about to be everywhere π
@ClaudeDevs practical unlock: hit an edge case mid-feature, /fork to investigate, keep building. fork returns with context while you never lost your train of thought. kills the 'open 3 claude tabs and track which is which' habit π₯