Ride to Rebuild is finished. 21 Foster City households rode 2,390.6 miles in September. $2,578.90 is going to the families displaced by the Sea Spray Lane fire, and it's already with the fund. Thank you to everyone who rode. muggle-ai.com/charity

Today is the last day to ride for Ride to Rebuild. 2,297.5 miles by 20 Foster City households so far — $2,485.80 unlocked for the Spinnaker Cove Fire Relief Fund. $514.20 of the $3,000 still unclaimed. $1 an adult mile. $10 a child's ride. muggle-ai.com/charity

IMPORTANT: If you are using Claude Works: SAVE EVERYTHING ON YOUR DISK!!!!! Plans, artifacts, decks, data, everything. No idea why people so hyped about Claude Works. It is such a sub-par product. Here is my experience: I tried oneshotting my sharing's deck with some ideas. The content it created is just...bad. And I ended up spending another day to handhold the agent, slide by slide, point by point, sentence by sentence. No way it can be more frustrating than this, right? ...The chat session disappeared after one day. No where to be found. Not phone client, not desktop client not even web client. Not in projects, not in code, not in chat. Just a lot of PR.

See how Muggle Test testing a new navigation bar feature for "Ride to Rebuild" campaign! ON MY DEV STATION!! Try us: muggletest.com/

CEO of Muggle AI here: My setup is Claude Code agent view remote session + Muggle Works agent harness + github pull request view. Works perfectly when I am out meeting with customers. - With Muggle Work harness, coding agent can loop full dev cycle - design, code, unit test, end-to-end test, PR and comments (thx to Muggle Test) in Claude Code, until PR completed. - Any question or comment on the PR (via Github UI) will get addressed and replied by the same Claude Code agent, with the full context. (you can shift the design with comments too) - PR shows you step-by-step screenshots of end-to-end test of all impacted test cases. How cool is that! Try us: muggletest.com See my profile highlight for demo: x.com/Muggle_AI/status/20614…

We're giving out 10 Muggle AI x MIIR coffee mugs to our early users! If you haven't already, register your account at muggletest.com. Comment and repost this. We'll randomly draw and contact you for your shipping address by 10/01. Try us: muggletest.com

I built "Ride to Rebuild" with Muggle Works harness + Muggle Test (End-to-end QA) in Claude Code. This is how it works on your desktop: 1. /mdo [describe what you want to build] 2. Confirm design details, and kick off implementation 3. Wait for harness spins up servers, and see E2E complete on you desktop, posting the result to PR 4. Review the PR evidence and comment on code if needed, which can kick off another round of /mdo

We have reached the $1,000 first donation milestone for the Sea Spray Lane families in just a week! Thank you for your support, thank you @SportsBasement redwood city, and FC neighbors! ❤️❤️❤️ Keep riding!

You can actually ask one Claude Code session to talk to another session. Here I am coordinating the local dev server across multiple tasks (each needs to spin up local services for Muggle Test e2e run). The session was holding it up after end-to-end test concluded. This is a very good feature.

Transform your PR workflow with Muggle Works + Muggle Test. We built our charity website in just 2 days — with AI building, testing, and validating every PR. • Watch an AI agent use your feature in real time and show exactly what works — and what doesn’t — step by step. • Test beyond sign-in and payment walls using securely uploaded secrets. • Run test cases in sequence so data and prerequisites are properly set up before each test. Build fast. Test like a real user. Ship with confidence. Try it: muggletest.com

A huge thank you to everyone who has supported our Ride to Rebuild fundraiser! 🚴❤️ We’re only on Day 3, and we’ve already raised $300 — 30% of our first $1,000 fundraising goal! 🙌 And an amazing shoutout to our current leaderboard leader, who has already contributed 93.2 miles (150 km) of riding. That’s incredible! 💪💪💪 We’d love to see even more friends and families join us. Participating is super easy: Ride → Record → Upload a screenshot → Donate Just record your ride using Apple Watch, Strava, a bike computer, or any other cycling app, then upload a screenshot like the examples in photos 3, 4, and 5. Our AI automatically reads the ride details and prepares the report for you to review and confirm before the donation is submitted. You don’t need to ride far — every ride and every mile counts. Looking forward to seeing more of you join the challenge and more familiar names on the leaderboard! 🚴‍♀️🚴‍♂️ Ride together. Rebuild together. Every mile makes a difference. ❤️

Muggle AI and Lin&Wang family foundation are co-hosting “Ride to rebuild” fundraiser for those impacted by the Aug 2nd fire at Sea Spray Lane, in Foster City. Between 2026/09/01-2026/10/01, we will donate $1 for every mile adult ride on a bike, scooter, skate board or strider, as long as it is non-electric. $10 match for any minors’ ride regardless of mileages. Up to $3000, with $1000 guaranteed min donation. Simply upload a screenshot of your ride from Strava, Apple fitness, etc, you will see the mileage show up on your account and leaderboard: muggle-ai.com/charity/leader… How to start: muggle-ai.com/charity/how-it… For more details, check out the website for this event: muggle-ai.com/charity Question? Email: admin@linwangfamilyfoundation.org Technical Support: support@muggle-ai.com

If you ever "resolved a conflict" or "rebase to master" on git, this is another reason you need to try Muggle Work harness, it gives you 1. Auto track all active PRs in the chat session 2. Auto resolve merge conflict, rebase master, fix broken CI for you, with your design and chat in context. Keeping your PR clean and merge ready. 3. Comment on PR and let coding agent address it, then reply. 4. Optionally run a E2E test and attach the result screenshots to your PR. Try us today: muggletest.com/

Built our fundraiser website with Muggle Works harness + Muggle Test E2E testing agent, in one day. Try us: muggletest.com/

*me when Claude is down*

Have you ever needing to tweak your original design after a second thought, or after seeing the PR implementation, or just want to change a few presumptions? Muggle Works harness gives you the "pill of regret", by handling your PR comments properly via spec update, re-implement, and re-test, giving you a smooth end to end vibecoding experience. Try our open-source harness! In your Claude Code: --------------- /plugin marketplace add github.com/multiplex-ai/mugg… /plugin install muggleai@muggle-works /muggle:muggle-pr-followup --------------- Also check out Muggle Test (end-to-end QA agent): muggletest.com/

When you have 13 parallel PRs going, you will need to manage conflicts and rebasing of the other 12 PRs after *one* PR merged. Our open-sourced Muggle Works harness automate all these (github.com/multiplex-ai/mugg…) . Even better, Muggle Works address your PR comments by coordinating works among PRs, syncing design spec with the latest comments, and automatically fixes CI/conflicts with the full context of the agent session. Also optionally integrating with Muggle Test for E2E validations. To install: /plugin marketplace add github.com/multiplex-ai/mugg… /plugin install muggleai@muggle-works /muggle:muggle-pr-followup

Muggle Works harness + Muggle Test Agent is a super powerful combo to build new features. Harness manages loops until each milestones are reached - specs, build, test, and present the result (PR + end-to-end user flow report). Agent takes care of the heavy-lifting of acceptance testing: understand the feature, log in, navigate, and deliver end-to-end walkthrough of the new feature. ---------------------------- Muggle Test: muggletest.com/ Install Muggle Works Harness in Claude Code: /plugin marketplace add github.com/multiplex-ai/mugg… /plugin install muggleai@muggle-works /muggle-do "build me a new feature" ------------------------------------------------- (See screenshot how harness + detects a missing "accordion" to view properties during acceptance test, autonomously check the step #28 screenshot, and drive a new dev cycle to fix the UX problem, post the latest end-to-end screenshots to PR at the end)

We ran this prompt in Claude Code today, which instantly saved us $2K/month on our Azure bill. This is pretty amazing: "Reduce my Azure subscription costs"

Devs, If you agree PR review is a MUST while building with Claude Code, try use Muggle Works harness for an extremely smooth experience. This harness streamline your Claude coding flow into: 1. Describe requirements 2. Review, comment, merge a spec PR 3. Review, comment, merge implementation PRs, each comes with step-by-step walkthrough screenshots, even behind login wall. 4. Merged and run full E2E agent tests in CI (on all test cases) Try Muggle Test (muggletest.com) + Muggle Works in Claude Code: =============== /plugin marketplace add github.com/multiplex-ai/mugg… /plugin install muggleai@muggle-works ===============

Claude Code users, if you cannot find the chat that created your PRs, try installing Muggle Work harness. This harness make your PR process fully autonomous. Simply having installed the harness gives you: 1. Track PRs the chat created (screenshot #1) 2. Powerful QA agent auto E2E test (over login wall, pay wall) 3. Code agent auto reply and address your PR comments (screenshot #2) 4. Work on your sessions via github + claude code mobile (w/ Remote Control enabled) ==================== /plugin marketplace add github.com/multiplex-ai/mugg… /plugin install muggleai@muggle-works /muggle:muggle-pr-followup ====================

Are you also waiting for the limit to reset?

Wouldn't it be nice to recalibrate design according to your PR comments? Try our super smooth vibe coding experience in Claude Code (w/ Muggle works): ---------------- /plugin marketplace add github.com/multiplex-ai/mugg… /plugin install muggleai@muggle-works /muggle:muggle-pr-followup ----------------

This is how we use coding agent AI slop free. Muggle Works harness builds, tests and tracks multiple PRs at the same time. It can response to your comment with architecture design change, sync changes among repos, and reply every one of your PR comments with updates. Can your coding agent do the same? Keep your PRs on track today: -------------- /plugin marketplace add github.com/multiplex-ai/mugg… /plugin install muggleai@muggle-works /muggle-pr-followup --------------

Why is all my feed filled with cringed posts from fake AI account pretending to be startup founder... 😭

Founder has no vacation. Watch how I use Muggle Test with Claude Code/Github to ship 12 PRs during my flight to Montreal: Try our free agent harness: muggletest.com

We almost hit our first milestone of 10 paying customers! 🚀🚀🚀🚀🚀🚀🚀🚀 Use Muggle Test (muggletest.com) boost your AI coding productivity by getting a end-to-end usage report of your new features. DM me for free 1-month trial

The productivity boost with Muggle Works harness is real. Here is Muggle AI’s Github contribution grapg since 2025: 6 months into 2026 (Muggle Works launched in March), we have already surpassed total contributions in 2025 by 50%. Expect to deliver 3x in 2026 more than 2025. What is your produtivity boost with AI?

"What type of error case Muggle Test can catch?" Answer: Not only UI bug, but also broken E2E experience. Here is an example from a real PR: "a broken "create secret" api detected" Start your free trial on muggletest.com

Running /muggle-pr-followup keeps active sessions afloat on your phone Before: messy sessions everywhere in Claude Code Mobile App After: Active sessions keeps active and updated at the top

I want a better remote control experience in Claude Code 🥹🥹🥹🥹🥹🥹🥹

Do you want to vibe code on yiur phone like this, seeing what works what doesnt, use Muggle Test. Install our skill plugin at muggletest.com

What is up with all these random sessions on app from my Claude Code on server? There are so many of them and they keep jumping around.

Thank you @ClaudeDevs. Me seeing the weekly rising over the weekend:

I vibecode even when I am AFK too. You are one command away from this smooth mobile vibecoding experience. Try Muggle Test’s open-source agent harness FREE: Ask me anything in the comment below. /plugin marketplace add github.com/multiplex-ai/mugg… /plugin install muggleai@muggle-works

I need advise: What would you do if you are at 58% Claude Code weekly quota 2 days into the week?

You can tell how Muggle Test boosted Muggle AI's vibecoding productivity getting into 2026. Side-by-side Github contribution graphs within Muggle AI: Left #1: 2025 Code Contribution Graph Right #2: 2026 Code Contribution Graph Try use out: muggletest.com

Hi #Founder #buildinpublic #claudecodedev This is how I vibecode with Muggle Works harness on my IPhone when I am away from my desktop. - Enable remote control in Claude Code (/remote-control) on desktop - Run “/muggle-pr-followup” on desktop - All set! Grab your phone, go out, and start building Install: /plugin marketplace add github.com/multiplex-ai/mugg… /plugin install muggleai@muggle-works

Hi Founders! This is how I vibecode with Muggle Works harness on my IPhone when I am away from my desktop. - Enable remote control in Claude Code (/remote-control) on desktop - Run “/muggle-pr-followup” on desktop - All set! Grab your phone and start building Install: /plugin marketplace add github.com/multiplex-ai/mugg… /plugin install muggleai@muggle-works

This is insanely good. If you are having problem finding the Claude Code session that created your PR, or go back and forth between IDE/Github and Claude Code: 🔥🔥🔥Do this instead: After PR is out, run "/muggle-pr-followup". - Locate and listen to all PRs created in the session - Comment-triggered dev cycle: Trigger dev cycle automatically (with optional E2E test) - Every comment gets a reply, and a (optional) visual walkthrough of the change in PR . Install: /plugin marketplace add github.com/multiplex-ai/mugg… /plugin install muggleai@muggle-works

CALL MY MOM!!!! Max’ed out 3 Claude Code sessions in a day🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🔥🫡🫡🫡🫡🫡🫡

First maxed out session on Opus 4.8 within 2 hours. It is looking pretty promising. (vs. 4.7) Muggle Test Vibecoding Stream English: youtube.com/live/k7C9cs8O5K8… 中文:youtube.com/live/9n_PEn-4Piw…

Can you please show the actual # PR in Clude Code (red box) @ClaudeDevs @bcherny . Knowing the count is not very helpful. Getting harder everyday tracing back to which session created which PR

Today’s vibecoding session: youtube.com/live/9n_PEn-4Piw… ✅ Tokenmaxx’ed

I need one more follower to 100. Founders/PM/Dev please help! Thank you!

Thank you for your support, everyone!

2 more to 50. Let’s go

My daily work is to PUA Opus 4.7 until it delivers. What is your story? #startup #claude #opus #vibecoding

你 Vibe Code 一个 SaaS。 测试都过了，部署也成功。 然后用户说：你的支付一直过不去。 我们做了 Muggle Test —— 让 AI 像真实用户一样在浏览器里跑一遍你的产品。 每个 PR 都验过，每次部署都重放。 你 push 的代码，是真的能用。 youtu.be/X3AVHKxXW3w