CodeTapper

\n## AI Coding Is Great — If You Enjoy Being Chained to Your Desk\n\nCodeTapper is a mobile terminal app built for Claude Code and Codex users, and its pitch is refreshingly honest: AI coding isn't a one-shot command. It's a loop. You prompt, the AI thinks, it runs, it errors, you prompt again. Author @rwecho described the cycle on V2EX: \"让它改一版。跑一下。报错了。再让它修。继续等。继续切 terminal。继续看输出。不知不觉，一个下午就过去了。\" (Ask it to revise. Run it. Error. Ask it to fix. Keep waiting. Keep switching terminals. Keep watching output. Before you know it, the afternoon is gone.)\n\nThe more productive AI becomes at writing code, the more it demands your presence to babysit the process. That's the paradox CodeTapper tries to break.\n\n## The \"Thinking Mode\" Trap\n\nThe most interesting part of @rwecho's original post is his observation about AI coding tools' behavioral design:\n\n> 现在很多 AI Coding 工具的 \"thinking 模式\"，其实有点像游戏。它会不断给你一点反馈，让你一直盯着它。\n\nTranslation: \"The 'thinking mode' of many AI coding tools is basically like a game. It keeps feeding you little bits of feedback to keep you staring at it.\"\n\nThis is a sharp insight. Claude Code streams its reasoning token by token. Codex shows intermediate steps. This real-time feedback creates a dopamine loop — you feel productive, you feel in control — but the cost is that you're anchored to your screen. You're not waiting for the AI; the AI is waiting for you, and it's trained you to believe you need to be there the moment it finishes.\n\n## A Remote Terminal, But for a Real Problem\n\nThe technical approach is straightforward: CodeTapper is essentially a remote terminal optimized for AI coding workflows. Your Claude Code or Codex instance runs on a server, the phone receives real-time output via WebSocket, and when the AI needs your input — say, after an error — you get a notification, tap in a response, and put the phone down.\n\nFrom the V2EX discussion thread (44 replies, ~3,887 views over 38 days), several developers resonated with the use case. One common scenario: you kick off an AI task before leaving the house, and now you can check progress and intervene from your phone instead of worrying. The tool also exposes an uncomfortable question: why haven't Claude, OpenAI, or Cursor built proper async mobile interaction into their own products?\n\n## Who Should Use It (and Who Shouldn't)\n\nCodeTapper fits a specific profile: developers who run long, multi-turn AI coding sessions where each iteration involves waiting minutes for execution and review. If your daily workflow is \"AI writes code → tests run → errors surface → AI revises\" on repeat, this app lets you reclaim the waiting time.\n\nBut the limitations are real. First, you need to set up a server that can run Claude Code or Codex — not trivial for casual users. Second, typing complex prompts on a phone keyboard is painful; serious manual intervention still requires a laptop. Third, and most importantly, this is a band-aid on a product gap. If Anthropic or OpenAI ship native async notifications with mobile-friendly interaction tomorrow, CodeTapper's core value proposition evaporates.\n\n## The \"Gap Product\" That Hits a Nerve\n\nCodeTapper isn't technically groundbreaking. It's a remote terminal with a specific use case. But that's exactly what makes it interesting as an indie developer product. @rwecho identified a friction point that major AI coding tools have ignored: the assumption that you'll always be sitting at your desk when the AI needs you.\n\nIn the current AI coding boom, the big platforms are racing to make AI write better code. Nobody's asking: what happens when the AI is done and you're at the grocery store? CodeTapper is a small product filling a small gap — but it's a gap that every Claude Code power user feels every single day. That's the kind of problem independent developers are best positioned to solve."}