量化策略研究员

@ME_Observer_ Ironically, he thought everyone would believe it

⚡️🚨U.S. Secretary of Defense Hegseth declares victory : "Never in history has a country been defeated as Iran has... We wiped it off the face of the earth, and then it was defeated

@chaomeigu_com 生不逢时

我不理解，之前字节跳动的 AI 手机，能点外卖，能发微信，能网购比价，能操作手机上的一切，才是普通人龙虾的最终解吧，为什么啥也操作不了的 OpenClaw 却成了最优解？🤔

@perplexity_ai Is there anyone interested in the local large model of lobster?

Announcing Personal Computer. Personal Computer is an always on, local merge with Perplexity Computer that works for you 24/7. It's personal, secure, and works across your files, apps, and sessions through a continuously running Mac mini.

@francistamhk @adrgrondin It is very necessary

@adrgrondin can we have internet fetching feature for the llm pls?

The latest update of Locally AI is out, bringing major performance improvements for the UI. Also Qwen 3.5 should be faster!

@Smitpatel_1306 @hwchase17 Less is more

@hwchase17 the rarest skill now is knowing what not to build.

@FLMdongtianfudi 有一本书有很系统的讲过

刚刚刷到影视飓风Tim的分享，他说了一个做内容的秘诀，把我给听愣了，但回过味来又觉得“太对了！” 他说，想创意的时候，千万不要一上来就追求线性、有条理。 人的大脑天生就不擅长这个。 正确的姿势是：先疯狂地、不分类地把所有零散的想法，哪怕再不靠谱，全都扔到一个乱七八糟的思维导图里1。就像打游戏放大招一样，先把所有技能都丢出去，直到感觉自己“山穷水尽”了。 然后再跳出来，像个局外人一样，审视这些乱七八糟的点子，给它们分类，找到哪个是核心观点，哪个是表现手法。 这感觉就像什么呢？先把一大堆乐高零件全都倒在地板上，看个过瘾，然后再慢慢地找零件，拼成你想要的模型。而不是一开始就按说明书一步步来。 这招真的绝了，不仅是做视频，感觉写东西、做PPT、策划活动都能用上。下次再没灵感，就知道该怎么办了：先允许自己“乱七八糟”一会！

@UnslothAI qwen3.5-0.4b qwen3.5-2b qwen3.5-9b

@UnslothAI openclaw+ollama+qwen3.5

Learn how to run Qwen3.5 locally using Claude Code. Our guide shows you how to run Qwen3.5 on your server for local agentic coding. We then build a Qwen 3.5 agent that autonomously fine-tunes models using Unsloth. Works on 24GB RAM or less. Guide: unsloth.ai/docs/basics/cl…

@fahdmirza thx

💥 Qwen3.5 0.8B vs 2B vs 4B vs 9B — Which One Should You Use and When? 🌀 ♠ Four models, one family — and one of them beats GPT-4o overall 🚀 🔹 Complete benchmark breakdown across 15 categories 🔹 Head to head matchups — where the gaps are real and where they are not 🔹 4B beats GPT-4o in 5 out of 6 categories — judged by Claude Opus 4.6 🔹 Thinking mode ON vs OFF — why it matters more than parameter count 🔹 Exact VRAM requirements for every model 🔹 Clear decision guide — which model for which hardware and task 🔹 All four are Apache 2.0, fully local, multimodal, 201 languages 🔥 Watch the full breakdown below 👇

@laozhang2579 打屁股

GPT-5.4物理效果真是炸裂💥 提示词我放👇

BTC 71000 已逆转趋势

@justinsuntron @totheskyvv 些几

@totheskyvv 东莞30年基业！

AI时代全球数字游民候选城市 惠州必须要吹爆 -安全背靠东大，完胜迪拜 -房价/房租/物价极低，完胜曼谷新山 -交通飞机/高铁直通深圳/香港，半小时经济圈 -气候四季如春，完胜纽约伦敦 -有全广东最好的海滩 -有山姆超市 -有与杭州一个级别的西湖 -空气质量全球前列 -罗浮山远超太平山 -14年后继承东莞30年基业 -孙哥气运之地 AI时代，顶峰相见！

@totheskyvv @justinsuntron 叼之别

@justinsuntron 还有媲美东莞的，千山万水，不如淡水

@justinsuntron 去惠州淡水叼之别

@somi_ai @JulianGoldieSEO 我尝试了两周，本地的真的很笨，甚至不能用。

@JulianGoldieSEO oh nice, been wanting to try this. how does it handle longer multi-step tasks though? that's usually where local models start to break down

OpenClaw + Qwen 3.5 just unlocked “Claude-level” local agents. No API bills, no cloud, no data leaving your machine. Here’s the setup flow. → Install the Ollama pre-release (needed for the newest Qwen build). → Pull Qwen 3.5 locally (it’s a big download). → Launch OpenClaw with the Qwen model. → Run one real workflow: publish, message, schedule. If you can run it locally, you own the whole stack. Save this video, you’ll set up a free private agent. Want the SOP? DM me. 💬