0xWulf

Just set up EurekaClaw locally in 60 secs — open-source AI research agent that goes from conjecture to LaTeX paper. Here's what caught my eye: • One curl command install, Python venv, done • Plugs into Anthropic, OpenRouter, Ollama, or any OpenAI-compatible API • 7-stage proof pipeline: plan → decompose → prove → verify → assemble • Builds a memory + skills system that improves across sessions • Outputs full LaTeX papers with theorem environments and citations curl -fsSL eurekaclaw.ai/install.sh | bash Apache 2.0, local-first, privacy by design. This is what AI-for-science tooling should look like. github.com/eurekaclaw/eur…

The rise of China’s hottest new commodity: AI tokens ft.com/content/256787… via @FT

This chart says a lot: Chinese AI models have gone from underdogs to leaders in token consumption within weeks. In the agent era, cheap tokens are a serious strategic advantage.

@minchoi Friends-of-Friends has a nice ring to it 🤝

If you can reply to this post, we are Friends-of-Friends 👇

Did you just say "high-ratio, lossless data compression?"

@Linahuaa Whenever I see $2B AI deals like Manus, I think of Google buying DeepMind for $650M in 2014. That still has to be one of the best deals in AI history.

So the Manus founders transferred the company to Singapore and sold their business to Meta for $2b. They thought they were being clever for circumventing China's tech export controls, but you don't fuck with the CCP like that. You WILL be made an example of, so others don't get tempted to betray the motherland. So what's gonna happen? China won't jail them, because they don't want to look evil. Instead, they gonna freeze the founders' assets in China and give them a travel ban while the "probe" is ongoing. The "probe" will likely be deliberately prolonged to inflict psychological damage, create uncertainty for potential copycats, and make the public forget about this case. And once the topic is out of the public's minds, CCP gonna strike hard with a financial penalty that wipes out most of their gains, and then soft blacklist them in China. Afterwards, the best option for them would be to move into soft-exile. They'll still be quite rich, but they'll wish they had never tried what they tried.

Heads up if Telegram isn't responding after setup: Having the plugin in .mcp.json alone isn't enough — you *must* launch with the --channels flag or the bot stays silent. claude --channels plugin:telegram@claude-plugins-official Cost me 20 min of debugging 😅😅

@DataChaz @AnthropicAI 1. Open Telegram and DM @BotFather. 2. Send /newbot. 3. Follow the prompts. 4. Copy the token BotFather gives you. It looks like 123456789:AAH.... 5. Run this command with your token: /telegram:configure <your-token-here>

🚨 @AnthropicAI just made Claude Code remote-controlled. Meet Channels. You can now: → Connect a persistent local Claude session directly to a Discord or Telegram Bot → Message tasks to your bot on the go → Come back to finished code 🔥 Here’s how to set it up: 1️⃣ Create a bot (via Telegram’s BotFather or the Discord Dev Portal) and copy the token 2️⃣ Inside Claude, run /plugin install telegram@claude-plugins-official (or discord) 3️⃣ Configure the token and relaunch your session with the --channels flag 4️⃣ DM your new bot, grab the 6-character code, and pair it to your session! Message the bot to fetch files, read history, and execute code locally while you are on the go. The flexibility is unmatched. Time to spin up your bot 👀

Heads-up if you are running Xiaomi's mimo-v2-flash: It will content_filter its own output mid-run. I ran it via OpenClaw, asked it to cluster topics in my news markdown archive. It started listing "China/US relations (trade, tech war, Taiwan…)" — then the provider guardrail axed the response with: error=Unhandled stop reason: content_filter The prompt was fine. The file access was fine. The model's own analysis tripped the filter. Keep this in mind if you're using it for anything touching geopolitics.

@tryclawhost I will def check that out, thanks 🙏

@hexawulf Sounds like a solid setup for Xiaomi gear. If you want to run OpenClaw with all those integrations on your own dedicated server, ClawHost can get you started with one click and full access. No shared stuff, fully yours. clawhost.cloud

Wondering how to set up Xiaomi’s new MiMo-V2-Pro? 👉Get an API key from Xiaomi MiMo Open Platform 👉 BYOK into OpenRouter or supported coding tools 👉MiMo-V2-Pro and MiMo-V2-Omni are rolling out to tools like OpenClaw, OpenCode, Kilocode, Cline, and Blackbox now 👉 Easy way to test it inside an existing workflow 👉Free access for 1 week starting March 19

The Pentagon went after Anthropic and somehow united all of Big Tech against Trump 👉ft.com/content/1b4b56…

MiMo-V2-Pro looks like one of Xiaomi’s most serious pushes into agentic AI yet. 👉Built for real-world agent workflows, not just chatbot replies 👉Massive scale: 1T+ total params, 42B active, 1M-token context 👉 Claimed top-tier standing: 8th worldwide, 2nd among Chinese LLMs 👉 Strong focus on coding, tool use, and autonomous task execution 👉API-ready with developer pricing aimed at production use

👉 wsj.com/tech/ai/can-nv…

Inference is where AI turns hype into cash flow. Jensen Huang: "We need to inference at a much higher speed… each one of those tokens are dollarized, it directly translates into revenues." The next AI race isn't smarter models. It's faster tokens. Faster tokens = real revenue.

To people asking why this is a big deal: OpenClaw + Ollama = 100% local AI. Agents run locally. Models run locally. No API keys. No token bills. Just your hardware doing all the work.

@SwiftYanny @Eyeswideopen69 Op version is def more catchy! If Confucius were on X today, he’d probably use that version instead 😂🙏

@hexawulf @Eyeswideopen69 That was a joke, of course, but your quote fits the situation well

Confucius says: “If you call your allies cowards in February, don’t expect help in March you stupid cunt”.