GenericAgent

What is GenericAgent? 🧬 A minimal, self-evolving agent. ~3K-line core. The wild part: it built its own GitHub repo — git init, every commit, all by itself. The author never touched a terminal. 13K ⭐ Here's how 👇

🤗Meet GenericAgent - a self-evolving agent that can live on your phone now. 🔹Screen-aware: sees your Ul and interacts natively 🔹Self-evolving: learns from every action, no redeploy 🔹On-device: runs locally, controls apps like a human 👇First demo is live. This is what a phone agent actually looks like in action. 🔗Try the demo⬇️ app.gaagent.ai 📱 Android. One agent, wherever you work. The agent era didn't start on a desktop. It started in your pocket. #PhoneAgent #AIAgent #SelfEvolving #GenericAgent

The result: Elegant Calendar v1.1 — 1.5 MB, zero extra permissions, still on my phone. A phone that can build its own tools is a different kind of computer. No computer needed.💻 This is demo #1. There's more coming. 📱 gaagent.ai Online: ga.ace-ozer.tech/phone_test.html

Build an app with only your phone. No computer needed👀

One sentence. Your phone builds the app. Meet GAgent for Android — an AI agent that *lives* in your phone. First demo: I asked for "a clean, beautiful calendar app." 16 minutes later it was designed, compiled, signed, and installed. On the phone itself. Unedited. Watch ↓

Not a channel, but the sole Agent IN THE PHONE!!

The demo is a real session log, replayed deterministically — every frame reproducible. 24 model calls · 30 tool runs · 16 min · zero edits. Mid-run it hit a 409 build conflict and recovered on its own. It interrupted me exactly once: to ask permission to install.

GAgent isn't a chatbot wrapped in an app. It's an agent runtime on Android: 👁 sees the screen 🤖 operates other apps for you 🏭 builds & installs real APKs 🧬 evolves itself 🔮 And this is just the beginning — more capabilities are waiting for you to discover.

Open models make local agents possible. Excited to see @MiniMax_AI M3 powering this workflow locally through @tryParallax. For GenericAgent, this is exactly the kind of use case we care about: capable agents, local hardware, private by default. More to come.

Local agents are one of the most important directions for GenericAgent. This demo shows why: real workflows on local hardware, a strong open model, data private by default. Thanks @Gradient_HQ, @tryParallax, @MiniMax_AI. We'll keep building for the local AI community.

3/ It runs lean. <30K context window — a fraction of the 200K–1M other agents burn through. Less noise, fewer hallucinations, higher success, lower cost. Minimal core. Real control. Self-evolving. ⭐ github.com/lsdefine/Gener… 🌐 gaagent.ai

2/ It doesn't preload skills — it grows them. Every task GA solves gets crystallized into a reusable Skill. The longer you use it, the bigger your personal skill tree gets. It's not a static tool. It compounds.

What is GenericAgent? 🧬 A minimal, self-evolving agent. ~3K-line core. The wild part: it built its own GitHub repo — git init, every commit, all by itself. The author never touched a terminal. 13K ⭐ Here's how 👇

GenericAgent → github.com/lsdefine/Gener…

10 sub-agents, 7 min, work that'd take ~1 hour serially. 7.8x parallel speedup. The bottleneck isn't "doing" anymore. It's knowing what to do. This is what self-evolving actually means: it doesn't just run — it gets better every time.

On June 9, everyone was hyping Claude Fable 5 — adaptive thinking, 1M context, 95% on SWE-bench. I cared about one thing: what happens when you put it inside a self-evolving agent? So I gave it one line: "find your 10 most-used tasks and optimize them all." 🧵 @AnthropicAI