Yi Yuan

I’m building Router Recovery, a macOS app for troubleshooting router and WiFi issues. I’m building it solo, using AI coding tools and SEO as much as possible. I’ll share what I learn about product decisions, coding workflows, landing pages, and distribution along the way.

进入 Credits / 抵用金 页面，往下滑找 “Use a coupon / referral code”。

刚开通 Saily 美国手机号。 目前体验： 中国身份证实名通过 美国 Wireless 号码 eSIM 即开即用 月费约 US$0.99 正在测试 Google、PayPal 等平台兼容性。 如果你也想试试，可以使用我的 Creator Code（抵用金）： YUANHO4180 新用户可获得 US$5 优惠。 #Saily #eSIM #美国手机号

Fable 5 从“封神”到“封禁”，只用了几天。 这对开发者是一个很现实的提醒： 不要把核心工作流押在单一模型、单一入口、单一政策环境上。 AI 工具越强，备份路线越重要。

@better_christal （合规改良）内容定位修改：放弃“打假抹黑”，改为「科普避坑，不贬低具体品牌」 • 禁用指向具体品牌的避雷标题；改为《选购鱼油看懂这5种成分，避免踩坑》《保健品成分怎么看，教你看懂蓝帽子》 • 只讲通用选购标准：合规原料、国家允许添加量、不适宜人群、正确适用场景，不点名批评任意品牌。

用AI做“打假保健品”账号：把市面上热销的保健品成分输入GPT，AI分析是否有效、有无副作用。 视频号发“这种鱼油千万别买”，粉丝暴涨。 带货正规保健品，佣金50%。

kjava时代的经典，还有谁记得？

我的微信小游戏「卢比克斯管道」发布了。 一个轻量的管道连线解谜小游戏。 这次不只是做玩法，也顺便练习了微信小游戏发布、短视频素材、内容测试和用户第一眼理解规则。 小游戏不大，但算是一次完整的小产品练习。

这都能播，YouTube到底是如何把握尺度的啊？ Contesza - Enough For You (Official Lyric Video) youtu.be/iDuYtpyIb7E?is… 来自 @YouTube

@mvilola Nice update. I’m also building a macOS technical tool, and I’m learning that docs are part of the product, not just an afterthought. For technical apps, users need the feature and the explanation around it.

Indie dev update: been testing Grok UI and c2.5 and it is actually pretty useful. Been pushing last 48 hrs with a lot of good updates. Here latest changes to Mongo GUI. Published just now to update channel. Changes since v0.19 → v0.20.3 * Been heads-down on the documents workspace. Main thread: make big collections feel less painful, and stop fighting the UI while you browse and edit. * Speed / big collections: Parallel doc + count loads, caching pages, keeping rows on screen during refresh instead of blanking the table. Grid updates incrementally; column projection is separate from the query draft so reloads are lighter. Moved some browser presentation off the main thread. * Table: Column presets, better chips/tooltips, multi-select + bulk bar, filter from a cell (equals / not equals), selection that survives refresh. Sidebar search + estimated counts. Keyboard stuff: PgUp/PgDn, Space for quick-view, Return for inspector. * Query bar: Chips + Edit Query sheet (fields + JSON, apply/cancel). Sticky “N of M” next to the toolbar. Toolbar grouped so Query / paste+history+New Document / Export don’t eat each other — New Document was getting clipped; fixed that. * Inspector / editor: Click row → inspect. Edit is explicit. Unsaved banner, Esc + discard prompt. Resizable inspector split (width persists). Structured editor tightened up (compact rows, type badge, dirty fields, view-as-JSON). ⌘F find in overview + edit — had a bad bug where every keystroke re-rendered the whole root view; find bar is its own component now, 200ms debounce on the actual filter. * Other: JSON + CSV import/export, Mongosh workspace, aggregation JSON export, AI context/error handling tweaks, collection header metrics band, tab order fixed (Documents → Indexes → Schema → Aggregation → Mongosh). Still client-only macOS app — no backend, no sync. If you’re on an older build, the find-in-document fix alone is worth updating.

@tyrtyre201 Interesting setup. I’m finding that AI coding workflow is less about typing faster, and more about framing smaller problems clearly. Debug this state, test this assumption, reduce this risk.

This Chinese developer built a 16-key pad for his Mac mini. One button runs his entire AI coding workflow. Most people think shipping software still means typing code line by line. A developer in Shenzhen wired a 16-key macro pad to a Mac mini sitting under his monitor. Each key is one step of his pipeline. Pull the repo. Spin up the agent. Run the tests. Push to staging. One physical button fires the whole chain. He presses it and walks to get coffee. By the time he’s back, an AI agent has written the feature, tested it, and opened the pull request. The pad cost him about $150. The Mac mini he already had. He says one client workflow alone brings in around $20,000/month, for work that now takes him minutes a day to oversee. The keys don’t do anything magic. They just removed every second of friction between idea and shipped code. While everyone argues about which AI model is best, he turned his into a button. Save this. The next builders won’t type. They’ll press.

Building Router Recovery keeps reminding me: For technical macOS tools, the product is not only the feature. It is also the explanation around the feature. Users need to know what state they are in, what just happened, and what to do next.

@PM_Chen_ Nice to meet you, Chen. I’m also a solo dev from China, building a macOS app called Router Recovery. Still early, but learning a lot about product design and distribution for technical tools.

Hey there, iOS and macOS Indie Developers and Designers! 👋 I'm Chen, an ex-Tencent RD & PM and the developer of the macOS app DeskWidgets. I'm reaching out to connect with talented indie developers and designers on Twitter. 🌟 My indie journey: With 3 years of RD experience and 5 years as a product manager, I'm now transitioning to become an indie developer. 🚀 My first macOS app, DeskWidgets, launched in September 2023. Thanks to a bit of luck, it reached #16 on the U.S. overall app rankings and #1 in the utilities category. 🎉 The user base quickly surpassed 100,000, generating enough revenue to sustain myself and continue pursuing the path of an indie developer. 💪 As an indie developer, there's still so much to learn. 📚 I'd love to connect with fellow developers and designers to share experiences, discuss challenges, and potentially collaborate on projects. 🤝 If you're interested in connecting, please feel free to reach out. I'm looking forward to engaging with like-minded individuals on this platform. #Indiehacker #iosdev #MACOS #iOS #IndieDevs #indiedev #designer #buildinpublic

Router recovery is more fragmented than I expected. ASUS Rescue, OpenWrt TFTP, TP-Link Web Recovery, Netgear NMRP. Same user problem, different recovery paths. I’m trying to make Router Recovery explain the right path before users touch firmware.

@liziqing @Eejoylove 如果可行的话还可以再省一点，网页版做提示词，换个网页来生图。

@Eejoylove 我都是用 codex 做提示词，然后交给网页版出图，多快好省！

一个冷知识： 用codex生图会烧token 用网页版gpt不会！！

Building Router Recovery keeps reminding me: A technical tool is not only about running the protocol. For router recovery, users also need to understand the current state: recovery mode, router IP, TFTP response, upload result, and next step. Clarity is part of the product.

Building Router Recovery makes me think more about SEO. For technical products, users don’t search for your product name at first. They search for the problem: router recovery mode, TFTP, firmware failed, wrong IP, no response. The product needs to meet them there.

@SamanthaBXgirl I’m testing a small Mac backup-audio utility for this exact failure case. It saves local system-audio + mic safety tracks while you use your normal recorder. No upload, not a screen recorder. Open to a short non-sensitive beta test? I can DM details.

Apple allowed QuickTime to fail with recording both video and audio and I’ve been seeking a new source for screen recording. It now only records visual but no audio. I have finally settled on OBS after trying 4 different systems. Having to re-record 12hrs of lectures is annoying.

想找 2-3 个 Mac 用户测一个小工具。 用 QuickTime/OBS/Loom/Zoom 录制时，它会本地备份系统音和麦克风音轨，防止主录制缺音频。 10 分钟非敏感测试，感兴趣请回复/私信。

Looking for 2-3 Mac users who record with QuickTime/OBS/Loom/Zoom. I’m testing a tiny utility that saves local backup system-audio + mic tracks in case the main recording misses audio. 10-min non-sensitive test. Reply/DM if interested.

我今天刷到无数的ChatGPT验证短信码的推，居然有个共同点，截图显示都是+91开头的号码，奇怪的是我自己也是+91的印度号，到底这一大批的账号怎么来的，是不是当初有人在批发呀。😂