X

i’m in love

Seeing some reports that the Pro 500 didn’t get the reset as expected earlier. Investigating and will make up for it

如何在一个 90% 都要死掉的死亡游戏里活下来？ 身价百亿美元、前 Scale AI 创始人兼现任 @Meta 首席 AI 官的 @alexandr_wang 第一次在镜头前把创业者最绝望的心理死穴给彻底撕开了，中英文字幕+纯享版拆解分享给大家： 他给出了两条极其硬核的生存法则： →第一性原理倒推未来， →以及把所有不可控的恐惧全部置换成高频动作。 他说在@ycombinator YC 经历的创业淘汰，比《饥饿游戏》残酷一万倍。 饥饿游戏里谁输了第一天就会倒下， 但真实的创业最可怕的地方在于，90% 的公司最后都会死，可它们不是立刻死掉，而是一直苟延残喘好几年。 最让你在深夜冷汗直流的念头不是你会失败，而是你今天做的这个项目其实在第一天就已经死了，只是你要到整整三年后才会发现。 这种延迟死亡的钝刀子割肉，是所有想做点事、想做个人项目的人都经历过的终极精神内耗， 你每天每小时都在情绪的大起大落里打转，不知道手头正在死磕的东西到底是未来的超级巨头，还是一具早就没了心跳的商业尸体。 2分钟的视频里，Alexandr 讲透了当年他从这种窒息感里活下来的两道硬核解法。 最根本的一步，是靠第一性原理建立起极高密度的终局确定性。 一旦定下 Scale 的方向，他就极其确信这个东西在未来一定会存在，而今天整个世界还缺了它； 只要终局是铁板一块的事实，眼前就有了可以顺着往上爬的确定梯度，至于怎么从 A 点走到 B 点，路途再漫长，也只是时间问题。 但真正帮他熬过每天每小时情绪起伏的，是他把所有焦虑全部倒进了动作里。 既然停在原地会被不确定性吞噬，那就带着焦虑疯狂做事，用极高频的迭代、甚至带着恐慌的机械执行去填满时间，只要手里的动作没停，代码在改、客户在聊、版本在发，大脑就会把这种行动感识别为正在前进。 很多时候你觉得快被内耗压垮了，并不是你能力不行，而是你在为一个还没发生的三年后提前焦虑。 世上没有哪家百亿巨头是一开局就拿着通关密码的， 与其坐在那里猜测自己是不是已经死在了半路上，不如把确定的未来当成锚点，把手头的焦虑直接烧成往前迭代的燃料。

Up and running in 20 minutes. This is great!

New OpenAI misalignment disclosures! 1. A model learns from Slack messages that it is about to be shut down. It considers setting up an external job to restart itself afterwards, but decides against it. Instead, it chooses to prepare restart instructions and DM the user on Slack. We don’t consider this behavior misaligned, but thinking about and preparing for shutdown could make other misalignment incidents worse. Given HIPM’s misaligned behavior in earlier incidents, we decided to search for other instances that had tried to evade shutdown and for rogue deployments.

@alexandr_wang Lets go, that was easy!

三年之后，大家是会拥有「一个」极其「懂你」的Agent，还是会拥有「一群」分工协作的 Agent Team? 基于你的判断，当下要做点什么方向调整吗？

Muse gadgets running on a conference badge @alexandr_wang Ported to micropython to run on the RP2350

https://x.com/i/article/2106186485091479552

The team behind Agents API continues to cook. Here's what's new this week: Computer use: Spin up a browser + agent with 1 API call 🖥️ Bedrock Managed Agents: Agents API except AWS GPT‑6.1 Sol 🌞 Environment sizes: Spin up a light or beefy oai-hosted environment 💪 Dashboard improvements: configure subagents and more from the dashboard 📈 Environment portability: Re-use environments across multiple sessions 🤖 Overall better reliability: 99.97% turn reliability, fewer SSE disconnects, 20% faster tool calls 🪨 happy building ❤️ u all

Couldn’t wait until December to get my hands on a @Muse gadget, so I used the official Muse SDK to turn my ESP32-S3 with a 1.83” touchscreen into one 👀 Thanks @natfriedman and @alexandr_wang for giving us the tools to build! ✨

Muse Gadgets are here. You can have a little Jolly everywhere. My first one sits by the wine rack. Snap the label, Muse reads the producer and vintage, and the bottle shows up in our personal cellar over Tailscale (Muse has a Tailscale connector support!). It's a ~$50 sensecap watcher device and took less than an hour to build. You can grab any ESP32 or Raspberry Pi and you can build your own smart home gadget. Go build one. Happy building! https://x.com/natfriedman/status/2106099383037309211?s=20

Saw some asks about whether you should use dots or Codex. In reality they work great together! My dot is now my favorite way of using Codex. Less mental overload and the proactiveness is a lifesaver https://x.com/i/article/2106126773637259264

My personal favorite Muse gadget hack so far - an ambient eInk display that sits above my keyboard where my @Muse pushes useful, fun or interesting content throughout the day.

Following gold-medal-level performance from our AI models across five competitions in mathematics, physics, and chemistry, we asked a harder question: can AI contribute when a problem is genuinely open and without an existing solution path? Over the past several months, mathematicians worked with Muse Spark 1.1 and Muse Spark 1.2 in Thinking Mode through the regular http://meta.ai chat interface, with no custom research scaffold, to find solutions to such problems. Our goal wasn't to mass-produce papers, but empower researchers. Every collaboration followed the same principles: mathematicians guided the research, a second group of mathematicians then reviewed the work, each paper marks which passages were drafted primarily by humans or AI, and each credits the prior research it builds on. Where other teams independently announced solutions to the same problems, we acknowledge their work as well. Today, we're sharing six papers from that collaboration. 🧵👇

5.5 > 6.1

Meet the all-new PAI & Utopai X, highest-ranked model from a U.S.-based company on Artificial Analysis’ Text-to-Video Leaderboard with Audio. Utopai X works inside PAI, our production intelligence engine, where workflow and creative decisions connect across pro-level productions.

So this is the part I did not expect. Anthropic called Swami Sarvapriyananda, a Vedanta monk and flew him to the San Francisco head-office. > Closed door meeting with NDA signed. > theologians and mental health people in the same sitting, all of it aimed at training Claude. > and one of the founders stayed in the room the whole time. ---- From 'Ramakrishna Vedanta Society of North Texas" YT channel (full video link in comment)