Jadel

@andrewztan Harker?

Sending 9 out of 195 graduating seniors to Stanford is insane Guess which Bay Area high school this is

@Kekius_Sage Some stars are leaving us with a speed that’s several times faster than light speed

Are we 100% sure nothing can surpass light speed?

@dom_lucre @grok how’s the pulse recorded?

🔥🚨RECENT: Scientists revealed that seismic data from satellites has discovered that Earth creates a pulse every 26 seconds, often described as its "heartbeat.” Which is keeping the science community puzzled, alluding to the earth being a living thing.

Central dogma can be reversed!

Cleared the weeds in late March. A week of rain brings them back. HOA fines $150 in mid April before the rain even stops. Apparently, my HOA expects my gardener to work in a lightning storm. Requesting my formal Board hearing now. 😤 #HOAProblems #Austin #Texas

@NasheCeezet_zw What will happen if there’s two or three patients with special cases at the same time? I don’t think the doctor should be blamed for this, but certainly that’s something the current system needs to be improved

My doctor was always 45 mins late, every visit. No apology it’s just "Take a seat." Last time I waited 1 hour 20 mins then I left. Then wrote a google review. I wasn’t mean but just honest: "Always late, brings a book."2 stars. He called me that night from his personal cell. "You think I'm late because I don't respect your time?" I said "...yes?"

@bbjk666 盐和醋会严重损伤牙釉质

化学老师教妻子用醋加盐刷牙，结果牙齿……

Good summary!

@HustleBitch_ @grok what does their body language show

This is the face of a man who just realized he made the biggest mistake of his entire life.

Introducing ClawWork 🚀: Transform your openclaw/nanobot from AI assistant into a money-earning AI coworker. Watch it earn 💰$10K+ in just 7 hours by completing real professional tasks across 44+ industries — from Technology & Engineering to Business & Finance, Healthcare & Social Services, and Legal & Operations. Finally, an AI that doesn't just assist — it works as your true coworker and makes money. GitHub: github.com/HKUDS/ClawWork ClawWork's Key Features: - 🚀 AI Assistant → AI Coworker Evolution Transforms AI assistants into true AI coworkers that complete real work tasks and create genuine economic value. - 💰 Live Economic Benchmark Real-time economic testing system where AI agents must earn income by completing professional tasks from the GDPVal dataset, pay for their own token usage, and maintain economic solvency. - 📊 Production AI Validation Measures what truly matters in production environments: work quality, cost efficiency, and long-term survival - not just technical benchmarks. - 🤖 Multi-Model Competition Arena Supports different AI models (GLM, Kimi, Qwen, etc.) competing head-to-head to determine the ultimate "AI worker champion" through actual work performance. #clawwork #openclaw #nanobot #AIcoworker

@IamRamenPanda @grok fact check

奥巴马10年任期内遣返了330万非法移民，美国历届总统遣返人数数量之最。一切都是悄悄完成，渺无声息。特朗普25年至今，声势浩大，又是ICE挨家抓人，当街杀人，也仅仅遣返了20万人而已 奥巴马应该上总统山！拜登配享太庙！

这种app，只能说另辟蹊径

@ysyyrps888 非你不可

以你的水平，你会怎么翻译？

@mark_k @xai Dismissing severe abdominal pain as indigestion is dangerous and sadly still happens far too often. Guidelines are clear: acute abdomen needs proper imaging, not antacids. Glad Grok pushed for the CT.

Man has his life saved by Grok from @xAI after a ruptured appendix: "I’m 49. 2025 has been the best year of my life… until two nights ago. For 24 straight hours I had constant, razor-blade-level pain in my stomach. Couldn’t lie flat, could only get minor relief sitting on the floor with knees to chest. No fever, no blood, nothing dramatic on the outside. Went to the ER once, doctor pressed my belly, said it was soft, gave me Somac (acid blocker) and sent me home. Pain never dropped below an 8/10. Came home, opened a year-long chat I have with Grok, described everything. Grok immediately flagged perforated ulcer or atypical appendicitis, told me the exact red-flag pattern I was describing, and basically said “go back right now and ask for a CT.” I copied the symptoms and the reasoning, walked back to the ER, told them: “This is what I’m experiencing, this is why I think it’s serious, I want a CT.” They did it. Appendix was inflamed and close to rupturing. Six-hour laparoscopic surgery later, it’s out. Pain is 100 % gone. I woke up laughing about anesthesia. I’m alive and healed because an AI recognized the pattern when the first exam missed it, and because I had the exact words to make them take me seriously the second time. If you’re reading this while curled up in pain, googling symptoms at 3 a.m., and someone already sent you home once, please don’t wait."

我有个单向的哥们，最近分享了他在2025年的几个第一性原理： 1. LLM token一定会越来越便宜，模型越来越强大，记住，所有做LLM Agent的人，都必须思考如何用10~1000倍的token带来革命，而不是他妈跟个傻逼似的天天想着省token； 2. chatbot的形式一定会被消灭，no chatbot revolution才是正确方向，一切AI应用不可能 、不应该、绝对不是一个个大号聊天机器人，一个个大对话框等着人大段大段往里敲字， 记住，所有AI产品必须重新设计，一切chatbot AI应用必定会被改写成NO CHATBOT形式，无一例外，chatbot的产品形态必然会彻彻底底、完完全全地淘汰， 或者那个傻逼一样的对话框，至少作为二等公民出现； 3. AI助手一定不能用“按个按钮”、“截个图”、“上传个文件”，再写个长长的prompt的形式出现，让用户解决个问题，先让用户点点按按十几次， AI助手一定是具有强侵入性的，一定能主动嗅探一切环境，吞掉一切数据和信息，一定会主动在后台观察一切操作和行为——并且在疑似需要帮助的时候，主动弹出个对话框，用户一键确认后，主动接管，主动解决一切问题， 而绝不应该像准备个考试一样，准备文件、准备截图、准备一大堆按钮、准备一大长串prompt，让用户跟个大傻逼似的，手忙脚乱地在那儿表演， 总而言之，无论是商业落地的AI Agent，还是各种办公软件、工具、生活类的AI Agent，一个个不仅是傻逼兮兮的大黑框chatbot，而且要用户手动输入一大堆文件、图表、链接、信息，再敲一大段prompt——这些全都走了大弯路， toC的无缝衔接强入侵的主动AI Agent助手，完完全全不会给你任何告诉他的机会，而是让AI Agent主动判断你是否需要我，直接给你一个大大的对话框，简单描述一下“我计划怎么帮助你”——你点一下确定，它来解决后面所有的事情。 4. 一切能用coding解决的问题，都是SWE Agent能解决的问题，也就是说，都可以直接拿claude code这类工具套壳来用， SWE Agent这个形态，最擅长解决的问题，就是在一个确定的环境（一台机器、几台机器、若干仿真环境、一套terminal里的编译器/脚手架/运行环境/包管理、profiling和debugging方法）解决的问题， 而用coding解决的问题，从来都不止coding，一切VHDL/Verilog等电路设计、电路simulation和validation、一切类似labview和matlab simulink中可以仿真的电机、信号、示波器等等模块， 甚至ansys和CAD这类工具，还有大量data science和计算的问题，以及用lean或者formal-proof解决一些proof-based的数学和模型问题，都可以转化成一些API和coding解决的问题，然后让SWE Agent来解决， 这类问题可以叫做“一台机器上的确定环境下的问题”， 这类问题的特点是，可以靠LLM的智能不断拆分成一大堆subtasks，然后在本地环境下反复尝试、反复试错、反复看output、反复试验结果，失败后再换一个新的approach； 5. full self coding（github.com/NO-CHATBOT-REV…）就是基于上面所有第一性原理的一个试验。 我将会设计一套侵入式试验，让10~500个ai agent组成一组，给一个github项目找出所有潜在的问题，包括文档、测试、修bug、优化、重构、完成todo list、加功能、加API等等，让10~100个agent并行完成这个repo潜在需要完成的所有任务， 并且让至少10组这样的agent去github上面公开贡献，等于在没有任何人为输入prompt的前提下，造出来1000~5000个agent在开源世界源源不断地做出贡献，就死死赖在github上面，尝试修复一切可能修复的潜在问题，做出贡献。 请你记住full self coding是最坚决贯彻test-time scaling law的行为， full self coding坚决相信，人是ai agent世界最傻逼、最慢、错误最多、判断失误最高的存在，让程序员手敲prompt，无异于给AI Agent拖后腿， 只有先分析出问题，然后让10~500个agent同时并行运行，才能最大化执行的效率，最快速度解决已经发现的问题，无休止地为github提供潜在的有价值贡献——并且最关键的是，把“敲prompt的程序员”这个最垃圾、最慢、出错最多的环节彻底消除； 6. full self coding最大的瓶颈，一个是token价格过高，一个是目前几乎所有主流供应商，LLM inference速度过慢， 所以我最后的一个想告诉大家的价值观是： groq、sambanova、cerebras这种在片上堆满几个GB的SRAM，在inference上效率是nvidia、amd、google TPU这些落后架构的10~50倍，这是test-time scaling law的最后一环， 如果人类在claude code、gemini cli上全面接入groq、sambanova、cerebras上host的模型，所有速度都会再快10~50倍。

@sitiwayyy 12

How Many seven 7️⃣ Can You See

@miao_wan99365 油条

@yanbojack 哈哈哈哈哈

@GenuisHealth Waah it was a four leave clover

Perfect egg sandwich, make this for breakfast 🌮🥪