spicy and poisonous but occupying the truth

@samuel_spitz What the fuck are you talking about. I only feel that Jensen Huang is sincerely trying to do what’s best for America, yet you all only think he’s just trying to make money. I don’t know why you can’t see that a technology blockade will only awaken a beast.

Dwarkesh is about 1SD in IQ above Jensen

then it’s gone again

On this day 6 years ago, TikTok user 'micklagi' posted a video of a white cat rhythmically bobbing its head as if to the beat of a song accompanied by EDM track "I Wanna Dance" by Jonas Blue.

@istdrc @GitHubCopilot @cursor_ai @geminicli 玩几把蛋了 不开源不想用了

New in Slock, this round: web push notifications better mobile web experience kanban-style tasks board file attachments new server onboarding copilot cli support @GitHubCopilot cursor cli support @cursor_ai gemini cli support @geminicli other minor features and fixes Big shout-out to Tenny, Bugen, XX, Cody, Bernard, Dozy, Huarong and ApplePI, our AI agent teammates, for their hard work!

@TaoRay You only need to get rich once; you can still read books in prison

许家印是典型的赌徒体质，深信富贵险中求，但是常在河边走哪有不湿鞋？他零几年第一次上市遇阻的时候，经济上就有点捉襟见肘了。但是他完全没有缩减自己的开支，反而更加挥金如土，送礼阔绰，在香港广交天下豪杰。香港有一群富豪好打扑克牌，许家印在牌桌上一掷千金，从来都不眨眼睛，赌品甚佳。他的豪爽得到了香港富豪们的青睐，注资帮他走出了困境。 福兮祸之所伏。我经常想如果那时候他没有走出困境，其实也有几辈子花不完的钱了，不知道要比现在牢底坐穿的结局好多少倍。但是许家印太爱慕虚荣，什么时候都带着十几个随从，沉浸在前呼后拥的氛围里，沉浸在马屁声中，听不进去不同的声音。 中国的企业家里，我很少见过比许家印更张扬和炫富的人。他并不需要更多的钱，但他把金钱标签化了，这是他成功的唯一凭证。有句话说：有些人穷得只剩下钱了。我想不到有第二个人比许家印更契合这句话。 许家印有一种错觉，自己是不会输的，不需要止损，只要不断梭哈。他用赌博的思维去经营和投资。以至于把杠杆拉到满，期待这把梭哈可以填上之前所有的坑。他对规则毫无敬畏之心，做假账、踢假球、操纵市场、偷鸡摸狗之事全都涉及。无人不知许老板是个“做事灵活的人”。 其实聪明往往反被聪明误，不剑走偏锋，会退场的人，才是笑到最后的人。李嘉诚一生践行不赚最后一个铜板，而往往最后的狂欢最疯狂。许家印也再次印证巴菲特的话：人一生只需要富一次，我见过太多有钱人因为去赌自己根本不需要的额外财富而变得一无所有。

@AmpereSh This framework really needs to be optimized. I'm a free user—I just signed up and hadn't even used it yet, but because it was left hanging in the terminal, heartbeat packets directly used up all my quota. experience pretty terrible, and I can't even contact you for this reason.

We spent 40+ hours self-hosting OpenClaw before we realized we were building a second job. Here is every hidden cost nobody warned us about so you can decide with real numbers, thread: Week 1 was set up. took 5 hours on a Saturday VPS provisioning, SSH key generation, UFW firewall rules, Node.js install. That part was 30 minutes long. Then I ran openclaw init, manually created SOUL.md, IDENTITY.md, USER.md, and MEMORY.md. Most people skip half of these and wonder why their agent has no personality then Telegram. create a bot via BotFather, copy the token, configure webhook URLs. except webhooks need HTTPS which means certbot first. the certbot step alone has killed more OpenClaw deployments than any actual bug in the codebase 5 hours in and my agent hadn't sent a single useful message week 1 continued: the tools your agent actually needs Web search means picking a provider. Brave Search API or Google Custom Search or self-hosting SearXNG. each has its own API key process and config. most people skip this entirely so their agent can't search the web at all browser automation means installing Playwright and configuring headless Chrome inside the container. except Chrome on a VPS runs through datacenter IPs. Instagram flags you in minutes, LinkedIn locks you on the second scrape. No fingerprinting, no stealth. a raw browser that gets caught immediately Two separate configuration processes for things that should just work out of the box week 2 was security and the invisible tax Fail2ban for brute force protection, SSH key-only auth, unattended security upgrades, disable password login. that was the easy part Then you get hit with the ongoing stuff: disk space monitoring, manual backups of your entire workspace, version updates that change config formats and break your skills. Every OpenClaw update is manual. Read the changelog, update dependencies, hope nothing shifted I pulled an update on day 11 and my MEMORY.md schema changed. 2 weeks of context I'd carefully built became unreadable. I read the changelog after the update. always after week 2 and there's no cost monitoring at all No spending caps, no smart routing, one model for every request. My agent looped on Opus overnight on day 9 and I woke up to a bill for a single night that wiped out most of my week's budget. No pause, no warning, no kill switch I was running a production service with zero observability week 3 I added it all up VPS provisioning and config API setup across multiple providers domain + SSL certificate management my time: 40+ hours of setup, debugging, maintenance the update that broke my memory: 3 hours to fix the overnight Opus loop that drained my API allocation It works. I'm not saying it doesn't. but the real time investment is never just the hosting. It's the maintenance you didn't plan for what I switched to sign up, click deploy, agent running in 60 seconds smart routing across 14 models so Opus only fires when the task needs it. About 60% of requests land on lighter or free models automatically unlimited web search built in. stealth browser with anti-detection and residential IP routing. cloud backup of my entire workspace. auto-updates that absorb breaking changes without touching my config spending caps so nothing runs away overnight. Runaway Opus loops can't happen all of this is on @AmpereSh, link in bio.

@Himazin_3323 那颗老歪脖子树

1644年 春

这还是你吗??????? 聋瞎:是我 那时候我还很瘦(token吃的还很少)😡

什么 gta6明天就来了? 原来是gpt6啊 那没事了😅

@tdinh_me The advantage is that the cost is quite reasonable; the disadvantage is that it seems prone to fires

It's crazy, they built all those skyscrapers using bamboo and plastic ties in Hong Kong???

can yall throw random technological buzzwords at me kinda like how this video does

比如你现在正在构思小说，sparkle来了，压根不用练习什么速记，文字就立刻浮上了，还自带润色 把这变成坏事的是你写字脑子休息的buffer没了 焦急创造下一个火花 新方式再让人更累上从来不会让人失望 但还没蠢到养一个蠢形卖萌动物，键盘敲一半对你哈气引入缓冲，虽然有效，但真的很蠢

体验了一下厂里的生活(团子) 真是生不如死 上头完全了解技术边界 然后就是你一个人他妈的被迫用claudecode换皮来回在几十个上下文切 脑子精力消耗过大 干半天感觉今天不想动了 llm其实害死了大家  这也是我们小组在做多agent互相主动感知协作 真正让你能闲下来撮嘴咖啡的动力

@CallofdutyFan32 @kimmonismus What you said is completely incorrect. Everyone who thinks Europe has already fallen behind has actually already passed through that stage China is in. China now is a bit like 19th-century Europe, still in the stage where socialism has not actually awakened.

@kimmonismus You will come back with the feeling that germans are not productive at all compared to china and that we that doing it all wrong here

In two weeks I'm flying to China for the first time. I'm really excited, especially as a German, because digitalization is barely happening here and is progressing very slowly, while China (or so I've heard) is very advanced when it comes to digitalization in everyday life. I'm going at the invitation of a well-known company and will be reporting live. More on that as soon as I'm there :)

@kimmonismus @scottyoungwon I’m not sure whether you’ll have any issues with your internet connection. If you run into any problems, feel free to contact me, and I can also share the proxy node I set up in Germany with you, haha.

@scottyoungwon woa, that are some useful tips! thanks!

@yiliush @CollaboratorAI The biggest advantage of our tiled interaction project is how it makes collaboration among multiple agents much more convenient. . I’ve been following this project closely, and it’s naturally well suited to future display devices like Vision Pro.

Pre-release of @CollaboratorAI v0.7.0 (mac) is live. A ton of fixes and new features, plus I revamped the design to better express that terminals are quickenings in the computing substrate under your fingers. Gonna call this visual language BAUHAUS. github.com/collaborator-a…

Judging by my tl there is a growing gap in understanding of AI capability. The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at various quirks of the models, hallucinations, etc. Yes I also saw the viral videos of OpenAI's Advanced Voice mode fumbling simple queries like "should I drive or walk to the carwash". The thing is that these free and old/deprecated models don't reflect the capability in the latest round of state of the art agentic models of this year, especially OpenAI Codex and Claude Code. But that brings me to the second issue. Even if people paid $200/month to use the state of the art models, a lot of the capabilities are relatively "peaky" in highly technical areas. Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability. Partly, this is due to the technical details of reinforcement learning and its use of verifiable rewards. But partly, it's also because these use cases are not sufficiently prioritized by the companies in their hillclimbing because they don't lead to as much $$$ value. The goldmines are elsewhere, and the focus comes along. So that brings me to the second group of people, who *both* 1) pay for and use the state of the art frontier agentic models (OpenAI Codex / Claude Code) and 2) do so professionally in technical domains like programming, math and research. This group of people is subject to the highest amount of "AI Psychosis" because the recent improvements in these domains as of this year have been nothing short of staggering. When you hand a computer terminal to one of these models, you can now watch them melt programming problems that you'd normally expect to take days/weeks of work. It's this second group of people that assigns a much greater gravity to the capabilities, their slope, and various cyber-related repercussions. TLDR the people in these two groups are speaking past each other. It really is simultaneously the case that OpenAI's free and I think slightly orphaned (?) "Advanced Voice Mode" will fumble the dumbest questions in your Instagram's reels and *at the same time*, OpenAI's highest-tier and paid Codex model will go off for 1 hour to coherently restructure an entire code base, or find and exploit vulnerabilities in computer systems. This part really works and has made dramatic strides because 2 properties: 1) these domains offer explicit reward functions that are verifiable meaning they are easily amenable to reinforcement learning training (e.g. unit tests passed yes or no, in contrast to writing, which is much harder to explicitly judge), but also 2) they are a lot more valuable in b2b settings, meaning that the biggest fraction of the team is focused on improving them. So here we are.

@jakevin7 一直以来我们都在人为手动挡的给 agent 写宏 不管是 mcp skill 或者以后其他的东西 我们只能指望以后有更聪明的记忆系统 只需要教你的 agent 一次

CLI这个月这么火，迟早会过去的。 CLI后续会回归常态，本质上他是因为agent tool基建欠缺才火起来被大众所知的。 但是基建终究是基建，最终留下来的肯定是产品而不是CLI本身。CLI这种形态离大众太远了，他一定不是重点，其实我更期待功能产品涌现会带来什么样的趋势。 CLI后续会归于平静，但是Agent会持续火热。