AI is evolving. Humans are still buffering. Tech, robots, breakthroughs & digital chaos. Welcome to the future. It's weird here.

Chiyoda-ku, Tokyo
Joined November 2022
认准无糖可乐,定期定投标普500,你也是可乐的股东
16
andriod人生:知道为什么要去山姆会员购买采购食材,基本上在外面吃骨头,排骨没有想样的排骨样子和肉的样子,所以经常去买菜做饭的人都懂,去菜市场买肉绝对不会买边角料肉的!!!!外面包子馅就是其中一种
15
世界模型真正拼的不是画面,是连续动作里物理会不会穿帮,跟本问题是长时记忆能不能撑住一整段世界不塌。
Excited to share Genie 4 today our most realistic and interactive world model: - steer dynamic environments in real time, shaping events, camera angles, and actions - generate native audio matched to the world, from realistic background sounds to characters who can converse in real time! - simulate photorealistic scenes and real-world physics - create more coherent multi-step events through reasoning from Gemini We've started to roll out as a research preview to trusted testers. Coming soon to Google AI Ultra subscribers in Project Genie and early access partners via the Gemini API and AI Studio.
1
18
真正考验不是速度,是跨应用操作一出错谁来兜底,必需先把权限边界谈清楚再放手给它跑全流程。
GPT-6.1 Terra: delivering fast and capable computer use at a lower price. GPT-6.1 Terra understands what’s on screen, takes action across apps, writes naturally, and is a more capable model for your whole workday with the efficiency to take on more of it.
11
政府采购价打下来只是开头,各部门真正用起来之后数据合规和审计谁兜底才是硬仗,我在乙方项目里卡过这关,在谈细节就知道水有多深。
13
先说结论:Agent 这轮真正的分水岭不是模型,是组织能力 麦肯锡 2026 的调研数字很说明问题:大公司把 Agent 做到规模化的比例,一年内从 27% 升到 40%,小公司却停在 22% 同一份调研里还有个更狠的数字:32% 的受访组织因为可以自建,放弃了至少一款软件采购 1. 先对齐三组数字,再谈判断 昨天另一组数据:79% 的企业说自己在用 Agent,只有 11% 真正放进生产。Gartner 的预测是到 2026 年底,40% 的企业应用会内嵌 Agent,2025 年还不到 5%。BCG 则说企业 AI 支出已经翻倍到营收的 3.3%,到 2030 年 Agent 会占 AI 创造价值的 40% 数字口径各家不一样,但方向完全一致:用的人多,规模化的人少,买单逻辑在变 2. 大公司赢在哪 不是模型更好,是三件脏活有人干:数据能打通,效果能量化,出问题有人背锅。小公司卡住的地方恰恰是这三件,模型一换代,前面的试点就得重来 所以 40% 对 22% 的差距,本质是组织能力的差距,不是算力差距 3. 对卖软件的人意味着什么 32% 放弃采购这个信号,比任何模型发布都值钱。按座位卖 license 的生意正在被按结果卖 Agent 替代,客户不再为功能数买单,只为跑通的流程买单 做工具的独立开发者也一样:别再卖模型调用,卖一条跑通的工作流,越具体越值钱 4. 小公司和个人的打法 别跟大公司拼自建规模。正确姿势是买垂直场景的 Agent,或者把一个环节做到能被验收:省了多少小时,少了多少返工,数字摆出来再谈钱 风险提示先说清楚:以上都是调研机构的自报数据,规模化的定义各家不同,别当成精确到个位数的事实。但趋势本身已经够用了 一句话收束:Agent 的下半场,拼的不是谁先用上,是谁先把 Agent 写进自己的考核和流程里
12
卧槽,AI 今天干了件教科书级别的事 OpenAI 和 Anthropic 的模型各自拿出了指数时间假设(ETH)的反证,一个困扰学界快 30 年的复杂性猜想 关键不是模型多聪明,是验证方式:OpenAI 那份用 Lean 形式化证明走了一遍,还交给了包括原猜想提出者在内的外部团队复核 这事的潜台词很扎心:以后 AI 搞科研,瓶颈可能不是想不出来,而是人类验不验证得过来 几百篇建立在 ETH 之上的硬度结论,现在都得重新排队,今晚先看学界吵成什么样
10
还是太年轻了,悟性不够。我以为只有墙内网红经常犯的事
姆巴佩完全没意识到现在中国还是独裁的
33
照你这么说,每个明星轮流丑话一遍,那国内劳动力都是些啥,太不值钱呢,艹艹艹
东鹏特饮已经开始清库存了 因为姆巴佩丑化毛主席
1
79
听说深圳可以直连了吗,放假了吗,????有没深圳的朋友,最近有什么变化,深圳朋友总是喜欢搞钱,闷声发大财啊。。。。。
没想到,为迎接下个月的APEC会议。 ChatGPT的广告牌已经出现在深圳机场了。
16
用他自己的原话,很简单,普通人不会投资的人,你就买普尔500就行了,定投就好。。。。
段永平的福报来了。此前不断被段永平加仓押注的泡泡玛特,发布第三季度业绩公告,整体营收同比下降 45 - 50%,海外美洲同比下降 70 - 75%,公司已出现明显负增长。 虽然港股已经收盘,但按照币安、Bitget 等场外合约报价 16 美金,约 126 港元来算,段永平已浮亏达 24 亿港元,约 -16%。
9
长按 推荐tab 屏蔽视频 就可以了
我无法接受那个外星人科普的 AI 视频。。。每次都是皱着眉头滑走。。
1
69
给在国内公司上班的人建议: 在是同事面前不要分享如何使用claude或gpt deepseek,如何用ai提效任何奇巧,如自己的工作流 如果你的同事没有用过Claude,更不要帮他装!不要告诉他!如何注册 不要问为什么,这是在保护你自己,这是你最后一份工作,能拖就拖。
2
1
186
非 AI 股跑赢 6.4 个点 说明资金在重新订价 指数光环之外的公司才是真考验
BREAKING: The S&P 500 excluding AI stocks has outperformed the S&P 500 by +6.4 percentage points so far in October, on track for the best relative monthly performance on record. This marks a sharp reversal from underperformance of -5.1 percentage points in the prior month, the 2nd-worst on record. By comparison, the previous largest outperformance was recorded in July at +3.6 percentage points. Over the last 3 weeks alone, the S&P 500 ex-AI has outperformed the S&P 500 by +7.4 percentage points, the widest 3-week spread on record. So far in 2026, relative performance has been extremely volatile, with the S&P 500 ex-AI outperforming for 4 months and the S&P 500 for 6. Investors are cashing in massive profits from the AI trade.
1
50
电脑操作降价才是关键 门槛低了 企业才敢把日常流程郊出去试错
6.1 Terra is here! It is good at a lot of things, but we are especially excited for this new era of computer use.
10
先收高风险能力换准备时间是务实选择 防预体系跟不上的时候硬上只会把风险外溢
I am okay with Kimi 3.1 deliberately leaving out top-end offensive cyber capabilities from their training data. Discriminating against use cases is MUCH less bad than discriminating against users. And the world does need all the time it can get to prepare, harden, check old code, (properly) formally verify new code, etc before the full weight of very powerful cyber attacks is upon us. And if this first big wave of AI risk is not handled well, then we may well get a dystopia of highly locked down models, KYC gating on everything, highly intrusive surveillance on everything, a society where you may be a second or third or fourth-class citizen in terms of model access based on your credentials and nationality.
8
接口跑通才算数 权限和账本没理顺的智能提越多 返工的坑就越深
I have been thinking a lot about the rise of AI agents and how it will change the way we interact with internet products. It feels like we have officially left the Internet 2.0 era and are now in the take-off phase of Internet 3.0. The biggest change I see happening in our day-to-day lives is a lack of future interactions with multiple product interfaces. Rather than navigating to various websites and using these products like neanderthals, we are more likely to talk with our agents, then these agents will "use" the internet and various products. Everything becomes more digitally connected, including information and money, but we see a massive overhaul of the consumer internet experience at the same time. You can think of this as the internet plumbing getting more complex, but the interface getting simpler. This will be very obvious when it comes to software interfaces. I also see the same trend happening with how we interact with our computers and phones. The amount of typing done by the average worker will quickly contract the same way that writing with a pen has largely contracted as a percentage of overall communication. Speaking is faster than typing. It is easier and more natural. And it will unlock productivity, which drives economic returns for people and companies. We will look back and think it is ridiculous that we used to put our hands on a keyboard, type in a URL, navigate to a website, and click a bunch of buttons to accomplish something. You can see early signs of this shift towards simple, agentic workflows today. There is still a ton of opportunity ahead though. I would love to hear whether others are seeing these shifts in their own behavior over recent months?
8
这周 AI 圈的融资新闻,画风全变了。 Meta 在考虑最多950亿美元的股权融资扩建 AI 基建,xAI 和 SpaceX 则签了400亿美元的 Vera Rubin GPU 租赁。 一个靠卖股份拿钱,一个靠租卡拿算力,相同点是没人再靠讲故事融资了。 以前是先画饼再找钱,现在是先把资产负债表摊开再谈扩张,说明大家对这轮基建周期有多长,心里都没底。
11
说个大实话,OpenAI 这轮降价不是在卷模型,是在抢客户。 据 The Information 报道,Rippling 把写代码工作量里 OpenAI 的份额从16%提到52%,Anthropic 从57%掉到18%。 更狠的是价格:一家资产超70亿美元的美国银行,ChatGPT 席位翻倍到500人还没加钱,单人不到7美元,平时20美元起。 模型下半场比的不是谁更聪明,是谁先把大客户的账单抢过去。
10
是不是只有我的x APP 有问题啊,升级到最新版本 看视频没有声音啊,我的声音打开了,不是静音呢
7