🚀 "In the end, It's not about the model. It's about the human." Agent platform
@mulerun_ai is blowing up on Discord and X. Founder & Zhihu contributor DoG A 陈宇森 shares the story behind it.
💡 The name MuleRun came early.
Idea: Let AI mules handle boring tasks. Cursor is a 🥕 on the site — every click, a mule runs for you.
🧑🎨 Team's 70% are i-people. Abstract vibes, so the logo and UI are pixel art.
🧠 The core belief:
LLMs are strong, but far from AGI.
Productivity work is often just "understand context → use SaaS → make decisions." AI can already handle many SOP-style workflows today.
🤖 So before AGI arrives, let's use "LLM SOP" to engineer reliable, reusable, monetizable Agents — A high-certainty path forward.
🎮 Our first Agent? Runs Honkai: Star Rail dailies, like a human. Two Brazilian YouTubers posted videos and comments: "AI can do this now?" — Sometimes a small tool beats a big demo.
🌍 Why we built a platform, not just Agents:
The most powerful workflows live in people's heads, not public datasets.
Real know-how comes from lived experience.
💼 Who pays?
Professionals who save time.
HR, Ops, Freelancers, Recruiters...
If an Agent does 80% of the job & saves 50% of the time — that's real value.
🧑💻 Our play:
Let experts turn workflows into Agents.
Let pros use & pay for them.
→ Platform handles discovery, payments, infra then everyone wins.
🎯 In August-Sept beta:
A dev saw Nano Banana and made 5 visual Agents. Upload 1 pic → get a 3D model. A florist's wife prints them as photo cards — customers love them.
An engineer, tired of job hunting, built LinkedIn Match:
It reads your resume, finds jobs, and applies for you.
Saved hours per day.
A team with 10yrs of U.S. stock research built a public-facing Agent. One colleague used it to buy
$AAPL — made 12% gain.
🔑 These creators weren't chasing "stronger AI"
They just turned expertise into something useful.
Solving real pain > showing off models.
💸 For us, creator earnings = platform success.
Make an Agent → get listed → earn money → iterate.
No need to bring your own traffic. We handle that.
🧱 Under the hood:
• Model Layer: multi-LLM, RAG, tools
• Flow Layer: SOP engine (stable, observable)
• Execution Layer: interacts w/ real world
• Biz Layer: metering, billing, rev share
• Market Layer: templates, rankings, growth tools — core for creators
🧪 Still building, but 3 big goals:
• Rock-solid infra (SOP must not break)
• Viral use cases (save time or go viral)
• Creator flywheel (from long tail to pros)
📊 What we measure daily:
• Task success rates
• Median creator income
• User retention & repeat purchases
• Agent reuse & cross-scene value
🧠 We want "Agent natives": Not all coders. But people good at breaking down workflows & serving real needs. Someone who spots a need, builds a workflow, wraps it in an Agent, and shares it — That's the real magic.
👉 Let's MuleRun:
mulerun.com
📝 Full Post:
zhihu.com/question/195086602…
#AI #Agent #AGI #MuleRun #AgentPlatform