频谱 SPECTRUM只增不改 · 每句可核验RSS

刚刚电台

TERN RADIO

一手长访谈 · 刚刚,是我开口的时间,不是新闻的时间

潮线 TIDE · 语料至 2026-07-04 · 已发布 8

身心科学Mind & Body Science

一份被误读的研究、一场发布会,一夜之间让整个激素替代疗法归零。真正致命的从来不是药,是对研究的错误解读被媒体放大成头条。开药的医生当场就觉得不对——我的病人没多死于心脏病——可惜临床直觉敌不过发布会。二十年后同一批作者自己翻了案,代价是整整一代女性硬扛过来的绝经期。

Everyone was told, throw your hormones in the garbage. This is dangerous. What was crazy is those people who were prescribing hormone therapy, who were looking around saying, I don't understand. My patients aren't dying of heart disease, they're not getting extra breast cancers. Like this doesn't make any sense. And when people actually looked at the study, it didn't say any of those things.

「所有人都被告知:把你的激素药扔进垃圾桶,这很危险。荒唐的是,那些开激素疗法的医生环顾四周说,我不明白——我的病人没有死于心脏病,也没有多得乳腺癌,这根本说不通。而当人们真去翻那份研究时,它压根没说过那些话。」

The Diary Of A CEO with Steven Bartlett · Dr Rachel Rubin: Women’s Sexual Health, Menopause, Hormone Replacement Therapy (HRT), and Orgasms!长寿、营养与健康科学永久页 ↗
身心科学Mind & Body Science

问卷量出来的是刻板印象,速配现场量出来的才是欲望。二十年、四十多个国家,只要面对的是见过面的真人,男女对野心的偏好强度'完全相同'——所谓性别差异,大半活在你填表那一刻,不活在你心动那一刻。拿它当择偶铁律的人,测的其实是文化脚本,不是人本身。

But then when we flipped it and we looked at what the women were drawn to, not what they said, but what they were drawn to, they also tended to like the ambitious men a little bit, and the magnitude of that preference was identical.

「但当我们反过来看女性真正被什么吸引——不是她们嘴上说的,而是她们实际被吸引的——她们同样会稍微更偏爱有野心的男性,而且这份偏好的强度一模一样。」

Huberman Lab · Science of Attraction, Compatibility & Romance | Dr. Paul Eastwick关系、心理健康与自我改善永久页 ↗
人生策略、职业与个人表现Life strategy, career and personal performance

「我是不是不适合当 CEO」这个问题,最后不是被答案解决的,是被濒死取消的——生死关头没资格反刍。所以「创始人模式」是事后包装:先有一场差点死掉的意外,才有那套方法论。他现在赌 AI 时代这是唯一活法,等于把自己的事故开成处方,这才是这句话危险又诱人的地方。

So I don't think I was naturally a good CEO. And I remember up to the pandemic. At one point, I wonder, I wonder if I'm just not meant to be a CEO. Maybe I'm not meant to do it. And then the pandemic happened. We had a near death experience. And I said, well, screw all the rumination. Like, is do or die. And I feel like I learned how to do the job

「所以我不觉得我天生是个好 CEO。我记得一直到疫情之前,有一阵我在想,我是不是根本不适合当 CEO,也许我就不是干这块料的。然后疫情来了,公司经历了一次濒死。我说,去他的反刍内耗,这是生死关头。我觉得我就是那时候学会了怎么干这份工作」

Invest Like the Best · Brian Chesky - AI Founder Mode - [Invest Like the Best, EP.471]人生策略、职业与个人表现永久页 ↗
创始人与公司Founders & Companies

替罪羊的要件是不会还手,AI 恰好全占。这话从 Tobi 嘴里说出来分量不同:2022 年 Shopify 大裁员,他签的是「我赌错了疫情」,没把锅甩给任何技术。反过来听更冷:等真正的 AI 失业潮到来,狼来了喊过太多次,替罪羊用早了,真凶反而查不了。

What you will see is like AI is going to be blamed for absolutely everything, right? Like because it's the perfect scapegoat, it's the perfect gerardian scapegoat for everything, right? Like it's like it can't fight back

「你会看到,AI 将被拿来背所有的锅,对吧?因为它是完美的替罪羊,是吉拉尔意义上完美的替罪羊——它不会还手」

The Twenty Minute VC (20VC) · 20VC: Shopify's Tobi Lütke on How AI is a Scapegoat for Mass Layoffs & What Will Labour Markets Be in the Future | Why We Need More Scrutiny on Charitable Giving, Governments are…风投、融资与创业生态永久页 ↗
人生策略、职业与个人表现Life strategy, career and personal performance

把「相信直觉」拆成两半,是这句话的狠处:直觉感知的是需求,想法只是实现假设,胜率差三倍。Zynga 当年被骂克隆工厂,正是这套教义的代价——他宁要有纪律的抄加改,不要浪漫的四十五次乱枪。AI 把试错成本打到地板之后,这话只会更对。

we have these instincts like in our gut these human level instincts that are pretty much always right and then we, put these ideas on top of the instincts that are usually wrong, and my rule of thumb is your instincts are right ninety five percent of the time your ideas are wrong seventy five percent or at best right twenty five percent of the time

「我们内心有这些直觉,这种人性层面的直觉几乎总是对的,然后我们在直觉之上叠加想法,而想法通常是错的。我的经验法则是:你的直觉95%的时候是对的,你的想法75%是错的,最好的情况也只有25%是对的」

Lenny's Podcast: Product | Career | Growth · The hidden pattern behind successful products | Mark Pincus (founder of Zynga)人生策略、职业与个人表现永久页 ↗
AI 工程AI Engineering

最值钱的是最后半句:成事靠的不是更聪明,是“跑得够久”。Swarm 本质是给短上下文打的补丁——记不住才拆成十个脑子,可并行买来吞吐、丢掉共识:十个各跑一千步再开会,输给一个自己跑一万步。当单 agent 学会自己管上下文,时长就取代并行度,成了新的 scaling 维度。

We found a surprising success of like don't do a swarm or anything. Just have one Devin. You know, it does its own context mention. Just let it keep running for a while and give it some crazy tasks. I think we asked it to like rebuild like a Windows OS system. Yes. And it managed to do it just like you know going on for long enough.

「我们发现了一个出人意料的成功路径:别搞 swarm,什么都别搞,就用一个 Devin。它自己管理上下文,就让它一直跑下去,给它一些疯狂的任务。我们让它去重建一个类似 Windows 的操作系统。是的。它真的做成了——就是靠跑得够久。」

Latent Space: The AI Engineer Podcast · The Age of Async Agents — Cognition's Walden Yan & OpenInspect's Cole Murray企业 AI、Agent 与应用落地永久页 ↗
AI 工程AI Engineering

折旧牛熊之争的要害就在这三个词的错位:会计折旧、物理寿命、经济寿命。Balaban 赌第三个最长,但经济寿命不是工程常数,是市场变量——旧卡能活多久,取决于新卡缺不缺货。而“空头一直都错”这句话,从一个资产负债表上压满 GPU 的 CEO 嘴里说出来,本身就是一张仓位单。

But that's not the usable life. The usable life is longer than the accounting depreciation schedule. And what really matters is the economic usable life. And so what we're starting to see is that like the people who are the naysayers, oh, this is going to be, you're going to throw these GPUs out in five years, are completely wrong. They're completely wrong and they've been wrong the entire time.

「但那不是可用寿命。可用寿命比会计折旧表要长。而真正重要的是经济可用寿命。所以我们开始看到,那些唱衰的人——说什么这些 GPU 五年就得扔掉——完全错了。他们完全错了,而且从头到尾一直都错。」

The MAD Podcast with Matt Turck · The GPU Myth: State of AI Compute 2026 | Stephen BalabanAI 商业、科技市场与公司战略永久页 ↗
AI 工程AI Engineering

Transformer 赢,不是因为它聪明,是因为它没有结构——任何矩阵乘任何矩阵,才吃得下最大规模。"缩放律定义架构"这话反过来听才有意思:当战场从数据中心挪到手机和车机,功耗会重写缩放律,结构就会回来。Liquid AI 赌的是后半句。

I believe like scaling laws define architecture, you know how the how how does it do that? When you are talking about transformers being this revolutionary thing, we are talking about maximum scale.

「我相信,是缩放律在定义架构——它是怎么做到的?当你说 transformer 是革命性的东西时,你实际在说的是最大规模。」

The Cognitive Revolution · Intelligence on the Edge: Liquid AI's Ramin Hasani on the Search for Device-Native Foundation ModelsAI/ML 技术与研究前沿永久页 ↗