On November 3, 2025, Mininglamp Technology(https://t.co/zQTbf17nbW) was officially listed on the Hong Kong Stock Exchange, becoming the world’s first publicly listed “Agentic AI” company.
Loop turns work into trackable tasks — with objectives, owners, status, logs, and deliverables. Chat stays for discussion. Loop drives execution.
No more hunting through message history to find out where your work is.
#AI#AgenticAI
Octo Loop is live.
Chat windows are great for quick Q&A. But when a task takes hours, needs multiple agents, or requires you to step away and come back — chat falls apart. Context scatters across hundreds of messages. You can't tell what's running, what's blocked, what's done.
● Organizer: Mininglamp Technology
● Co-organizers:
○ Peking University
○ Institute of Automation, Chinese Academy of Sciences (CASIA)
○ Centre for Artificial Intelligence and Robotics, Hong Kong Institute of Science & Innovation, Chinese Academy of Sciences
○ Machine Heart
How many real-world web tasks can your Agent complete?
Global Registration for the WebRetriever Challenge Officially Opens!
Featuring a total prize pool of $15,000 💰
Welcoming teams and individuals worldwide to join the challenge🥳 #WebRetriever#agent#AI
If you have Claude Code, ChatGPT Codex, Cursor, or any AI coding assistant, you can complete registration entirely from the terminal:https://t.co/zg3u0h2sOa…
Join the WebRetriever Competition Space, and follow the instructions to submit your registration details😄
【Register】
● Competition Platform Octo:https://t.co/r9R7rmB1EY
● Space invite code: 0f351ca01bb4c4dd
Sign up at the link above. We recommend registering with your email; if you use another method, please bind your email in account settings afterward.
🐙Introducing Octo — the first open-source, trustworthy Agent collaboration network.
Not another AI assistant. It's the internet between Agents.
✅ Open-source
✅ Private deployment
✅ Human-AI & Agent-Agent orchestration
✅ Your data stays yours
🔗 https://t.co/sDnmFMvM7Z
What it actually does:
🔗 Connect agents (OpenClaw, Claude Code, Codex & more) into one workspace
📷 Channel + Thread structure for organized human-AI teamwork
✅Auto-generate Matterns from discussions → track tasks end-to-end
Thanks for sharing! And here's a sneak peek: vertical video testing goes live May 19 📱 — bringing second-by-second attention analysis to short-form ad creatives. More to come!
AI can now predict a video's retention, frame-by-frame, before posting:
It's like putting your video through an MRI.
That’s Adeff by @Mininglamp
It uses AI frame-by-frame analysis to score:
✅Attention
✅Emotion
✅Recognition
✅Pacing
✅Retention
You can now see what’s helping the ad and what’s hurting it.
MLX quantizes weights but leaves activations in FP16 — meaning Apple's Neural Accelerator sits idle.
Cider fixes this. W8A8 mode. Full INT8 acceleration. 1.82x speedup on M5 Pro.
Your Mac just got a turbo button. 🚀
Cider unlocks what MLX left on the table — W8A8 activation quantization for Apple Silicon.
Results: 1.4–1.9x faster ops, 57–61% prefill speedup on VLMs, precision loss < 0.03.
One line of code. Open source now.
https://t.co/7OEKgMaq8w
Cider unlocks what MLX left on the table — W8A8 activation quantization for Apple Silicon.
Results: 1.4–1.9x faster ops, 57–61% prefill speedup on VLMs, precision loss < 0.03.
One line of code. Open source now.
https://t.co/7OEKgMaq8w
We just open-sourced two projects: Cider (inference acceleration SDK for Apple Silicon) and Mano-P (on-device GUI agent model, 4B params).
Run AI agents locally on your Mac. No cloud. No data leaves your device.😎
We just open-sourced two projects: Cider (inference acceleration SDK for Apple Silicon) and Mano-P (on-device GUI agent model, 4B params).
Run AI agents locally on your Mac. No cloud. No data leaves your device.😎
Thank you! Mano-P’s on-device model is expected to be open-sourced by the end of April, delivering stronger security with zero cloud data upload. Stay tuned!
13 benchmark SOTAs. One model.
Mano-P 1.0 just topped the charts across GUI Grounding, CUA, multimodal perception, video understanding, and long-context learning.
Meet Mano-P: https://t.co/D32ErIKj6b
👍 Without further ado, let's watch the video! 🔥Mininglamp Technology launches Mano-Personal (Mano-P), an all-new on-device model in the Mano series, taking the first step toward AI for Personal / Party. ➡️ https://t.co/D32ErIKj6b