今日概览
8
节目总数
8
订阅源
8h50m
总时长
Top Story
【All for Today】🎙️ 每日播客精要全景图
首条总览
42章经
(1)
从蒸馏到合成数据到 RSI,模型竞争的下一个焦点是什么?|对谈 Evolvent AI 联创孟繁青
59m
2026/08/08
📝 AI
总结
.There's something there.🎼 やだ。🎼 我们最近聊了很多跟模型算法相关的,因为这块也非常热嘛,所以今天很开心请到U的AI的联创韩青,可以给大家先简单介绍一下。OK非常感谢曲老师的邀请,我是孟凡青,我是N博士,同时我也是U的AI这边的联创始人。然后我主要在我们这边是fo在CI这个rese的方向啊,我之前是在kimi有过一个比较长时间的实习经历。然后当时的话主要去做A的。😊还有L这块,然后主要是偏的方向。然后在ki的话主要的一些工作就包括像ki以及K2.5这样的一些重要模型的发版。那么从ki之后呢,我是跟呃一些朋友去我们就和伙创业,就是这个volAI这公司对于这个公司,我们目前的一个大的scope,就是说我们主要是fo在ICI这条线上。我觉得你的背景其实是很典型的。今天的一批。特别热的创始人,然后大家喜欢的华像对吧?就是非常年轻,然后PHD毕业,甚至还没有毕业就开始创业,而且是做的跟因为你之前在kiimmi做的也算posttrain嘛。我觉得这批人现在出来创业的非常多。所以你最近一年的包你创业算半嗯对半年差不多半年左右,你实际的感受怎么样?我会觉得其实在大厂跟在创业公司做事情会比较有不同。比方说一个很典型点,就是啊我在kiimmi其实主要是去。虚拟技术模型。但是事实上可能我到后期我会觉得说虚拟技术模型这件事其实在大厂内部它是一个相对比较工程化的事。那么可能我主要的工作也会去做一些核成数据,相当于因为其实整个一个训练啊,包括推的流程已经在内部是很稳定的了。那么我们可能来到这种创业公司以后,呃,我们肯定是就不训即模了,然后我们就需要找到一些与这些模型厂更正交的一个一些切入点,以及就是相当于我们需要。要去从零搭建一些像平台类的这种方面内部使用的一些infer。这个是一个比较大的不同。就是说我们可能并不是fo在推进模型治理,而是说我们是去做跟模型智力正交的,但是可能是直际应用更有用的东西。嗯,嗯我发现你做post training的人现在创业的是比较多的。嗯pre training的好像很少。对,这个其实很好理解。因为其实做pretrain那帮人,他们可能每天基本上打交道了要。就是data本身其实这块也比较少,因为培训data现在已经相当于比较饱和。那他们其实可能每天打交了大多数一些模型架构啊,包括是优化器这样的事。那么其实搞这种东西的话,它是相当是资源的一件事。所以来创业的话,初创公司也比较难cover这一块成本。就他做一些事情是在卡的技术是这样的,对吧那 training的人出来创业,你觉得他是比较擅长和适合的是做什么?从目前来看,整个 training它基本上是跟。一方面就是核人数据嘛,就核人数据是很重要的一块,就它相比于训练核成数据,我们对他要求是希望它更精细,就train的数据量不会像预训练那么大。所以这一块,那么还有一块其实就是做平台类的平台类可能听起来稍微比较抽象。其实我给他说白话就是做banch这个ban发展到现在,他已经不是说像我们可能24年25年看到的那种有很多的难题,让模型。去回答。而说这个办企呢,它本身就是比方说像porttrain办企这种,那么模型可能就是需要在他提供的这个环境里做非常长时间的一个探索去完成这样一个任务。那么现在A的办企很多都变成了这种。那么这种呢,它其实相比于我们刚刚说的人员的一个研究模式,它就更加更偏落地一点。因为其实就是当我们这个办级做出来以后,一般一个足够优秀的。对于A的企,他是说一个工程上实现就会比较好的。相对而言,他转成一个可以落地的商业模式也会更简单。所以我觉得这就是为啥这帮人,他们大多是搞A进的,还有企,他本身就是离这种落地更近的。所以他们这个创业也会更多。嗯,哎我觉得你说这个的白话还不够白啊,就是所以到底今天的ch你的跟之前的区别是什么?之前的ch我们可以理解为就是嗯类比的话就是做数学。题就是我有一个很难题,然后我让模型直接去回答,嗯,相当于他就是一个考生。但是这个考生他并没有他在日常生活中的一个环境。举个例子,比方说就白领,其实我们在日常工作中,我需要完成事,我不仅仅是靠我的大脑。之前办起大陆是只测模型大脑,我还需要去动手去操纵像飞书啊,像nion各种各样的软件。那么其实现在的办uch呢就相当于把这一块给他模拟出来了。比方说是白。白领的话,我们会建一个就是模拟白领这种工作软件的这样一个整体的环境,然后让A在里面去完成某些任务。那么这样的话,他就是不仅仅说我需要有做数学题那种强大推理能力,我还需要有我去跟这些工具进行交互的一个动手能力,最后我才能把这件事给做出来。哎,但你看之前的那个就是做题这件事情,包括这件事情,它的可验证性的,其实我理解是更强更简单的嗯对吧?就是你最后对一下结。我对就是对不对不对,那你说我建一个环境,最终怎么评判这个结果的好坏?这块其实就会相对于原来更复杂一点。因为其实原先的话,比方说我们出题,那首先我需要保证这个题是对的。现在的话其实也是类似的,就是当我们造了一个这种环境以后,我们需要确保就是我们造这个环境跟这个真实的环境是有烂的,一个比较简单的例子,就是说这个环境里我们造了一个notion。当然这世也里有一个真实notion。那么当同一个输入下,这两个输出是不是一样的。那如果一样的话,就可以证明我们造这个环境是对的嘛。那么还有一块就是呃我们出了题以及对于这个题配套的一个检查。就是现在来说的话,一般这个检查就不会像之前数学题那样那么简单,就是说可能就是一个判断是不是相等。那么现在检查它大多数都会分非常多的一个维度。就我们可以看到就是现在的。很多chmark像我们之前搞的cromark,或者是目前可能大家都在测了一些像automation bench这种,它的检查都是分为非常多的那种小的得分点了。那么这些小的得分点呢,目前是他还是比较去依赖于一个人的一个精细的设计。所以最后还是需要人来做这个检查嘛,对吧?其实也是人跟A一起去check了。嗯嗯明白,哎我最近听那一个说法就是有人会讲说post training的人,其实每天主要就是在做数据。确实因为就是像我刚刚说的,其实目前模型厂它对于整个training这块生态是很完备的。他就不像可能比方说这些大厂之外人,他如果他想去训练SFT一个模型。那他可能说我要去gethub上拉一个仓库。然后我要找到一些GPU,我要配一个环境,然后把我的数据整理好,然后去训一个模型,就是模型厂内部的话,他就是如果我一个研究员,我想去SF。模型我根本不需要这么复杂,因为相当于他把这一套生态给他做成了一个服务。这个服务的话就是说形式上就是我仅仅去提交一个请求,然后这个服务就会去处理我的请求,在远程去去SFT。然后我这边最后就拿到结果,相当于对我来说,就算我是一个什么也不会,我根本不知道SFT是什么来着。我要做的就是把这个数据做好,然后我就可以直接走完这一套完整流程。对,所以你会有。有讲说那post是不是其实相对是比较枯燥无聊的工作,就是认真的把数据做好啊,以及说那是不是现在所谓的这个训模型这件事,其实门槛没有大家想那么高嗯。其实对于后者我是认可的,就是去模型这件事的门槛确实是有很大程度降低的。当然这个是对大部分有消息来说,因为你比方说你要搞infr的话,其实相当于你是去构建那个服务的人,这块还是有比较多的坑的。但如果说是大部分普通的算法有消息,确实门槛是会降低的。但他到底是不是枯燥乏味呢,这件事反而还真不一定。因为其实所有东西按照他的一个发展的。看其实我们都会发现,不管是模型架构,它为什么最终到了conformer,还是说在RL的时候,为什么最终就过渡到了deep当提出了那个基于可验证的的那种R的方式。其实就是他会发现这个事物的发展规律一定是发展到越来越简单越来越直接面向结果最直接的这种形式。所以其实我觉得后训这块也是这样的,就是他也是发展到我可能不去关心具体算法上。面设计了一些trick,然而是直接关注最本质的就是给模型的数据这样一件事。所以这一块我觉得它发展是正常的。然后本身造数据的这个过程其实也是比较难的。确实是不同人,他就是能造出来数据是不一样的。并且的话我们最近也有一些工作,我们去做了一个SI batchch。然后我们在这个banch里,我们也去讨论了,就是agent他到底能不能直接完成去造数据,然后迭代起来去模型这件事。呃,就目前我们可以看出来,就是其实A键他还是只能做一些比较机械化的事。比方说我训练这一批数据,导致这个模型的回答的格式就错了。然后我后面发现A键要修改,他可以完成这样一件事,但他是无法提出一些造数据的所谓的inet的。呃,我们可以发现,其实现在模拟厂的研究员他在造数据上,他是有非常多的一些他们自己的no how的那这个其实也就是目前的一个核心能力吧。所以也。不能说这件事本身是枯燥乏味了。嗯,哎所以你看同样的人,他做出来的数据的结果不同,好坏大概体现在什么地方?嗯,有没有一个某一个具体的例子呀么其实这块目前来看他还是比较量化的。大概分为两个维度。因为目前的话造数据这件事一般也是针对给定的一些banuch。那么其实最简单的一个量化标准,就是说我在这个banch上到底能不能在没有任何作弊,或者说叫的情况下完成。这个提点就在这个banch mark上的效果可以提高哦,对,这个就是最直接的一件事。然后当然他可能也会有一些其他的考虑。比方说我这个模型,我在一些代码的班上,我去核人数据,然后去训练他,他会不会在一些比方说协作这种偏文科上面的事情上去有一些性能的降低。但你说的这个是结果上的评判。对对吧?对我就说是什么造成的不同的这个结果呢?主要就是他们核。出来那个数据的质量上,其实对于目前就是当我们没有去训的时候,我们去评判这个数据的质量,其实它还是相对比较难的。嗯,就是我们其实评判它,我们只能用一些量化指标去评判它。比方说这个数据的一个正确性。因为我们这个数据一般是说我们有一个问题,然后有一个模型轨迹,就会有一个答案嘛,那么其实一方面就是说我们看这个模型轨迹里到底有没有出现类似作弊这种情况。还有一方面是说对于。与我们合成的这份数据,它是不是一个难度是适中的。比方说如果太简单,数据肯定没有用。那么太难数据的话可定也学不到什么东西。所以我能不能换个角度理解,就是现在数据的质量,它仍然是一个比较难评判的事情,比较难。就是你最后还是得训了,就是结果好就是好不好就是不好,就在改。对,就是其实很多对于数据质量检测这种指标,它只能做一些像出筛,就正儿八经看它到底能不能work,还是。去训练。所以对于一个模型来讲,就是我们只说后训练部分,它的infr的好坏跟数据的好坏,你觉得这个重要性占比大概是怎么样?呃,目前来看是数据的好坏更重要。因为呃infr已经发展到一个比较稳定的程度。OK所以我们刚才讲的很多,我觉得是相对比较介绍型的,所以我觉得我们可以讲到说,那今天因为你正好也从kimi出来不久,对吧?然后现在kimi啊min max质什么的,都竞争的还蛮激烈的。大家最近。都在陆续的发行模型。嗯,所以你在你的视角看来,今天尤其是国内的这些模型,大概是发展到一个什么阶段,大家在拼的是什么东西?其实我觉得目前国内的模型就是相比于国外模型差距是一个在缩小的过程。嗯,比方说我们可以看到K3在国内这种卡可能只有国外10分之1的情况下,但是发展到了一个接近就是发布这种。当然就是你具体跟国外差多长时间这块其实也比较难估计。因为其实像国外那些厂他都会有自己一些存货,他可能没有发出来。但是从他们re出来看呢,结果是在接近的。那么我觉得这块它主要是来自于应该是主要是预训练这一块。就像我刚刚说的,国内的模型厂它有的一个计算资源是远远少于国外的,就是导致了他们需要用这些更少的资源怎么去完成一个比较好的后训练。那么这块他在国内我感觉基本上可以分成两种,一种就是就像kimi跟 deept,它会有一些架构上底层种创新。比方说像ki,它是一个线性注意力机制。。像有那个也是一种注意机制。就这种的话,它说白了就是相对于原始的那种ttion,它更省计算量。像这样的话,它就可以用更少线资源达到一个更好的效果。同时它也不会太损失性能以及推理速度。那...(已截断)
‘有撒好奇怪的’
(1)
我们追星追到“一无所有”:一个一个偶像都不外如此
1h 16m
2026/08/08
📝 AI
总结
Oh.Yeah.🎼 大家好大家好,好久不见。好久不见。😊🎼我是西Y,我是托利,欢迎收听今天的有赛好奇怪的。我们今天要聊,其实说了挺久了,就是上次就是要聊对对,就是因为你最近也是要去看个演唱会嘛。是的,然后你在那种呃粉丝的群里是吧?对,然刚好这两天这这这一个礼拜连续的那种热搜都跟很多都都是跟饭圈。跟追星有关的。,我们就先说昨天的热搜哎。就是你发给我的嘛,我就一下就觉得哇太痛心了。就是那个min娜的事情。是的,怎么讲呢?因为我们还是也是大的胆子来聊这个事儿。对,因为米娜的事情是因为网报,所以他呃经受不住这个压力。对,就是选择了比较极端的方式。对他确实也很极端,感觉他精神都崩溃了。对,嗯,就是直播直播。然后。哦,对吧?对,是直播在做几个极端的方式。然后如果我们现在聊万一如果我们的节目被人听到了,也很有可能会受到网曝,顺着网线都杀网线杀我们,对吧?嗯但是呃还是想还是很想聊,还是很想表达一下,就是对这件事情的一个看法看法吧。我觉得对我来讲不是看法,嗯,就是一种感受。嗯,就是因为这件。事情让我觉得说哎怎么讲,人还是要。珍惜自己吧,谁都不值得谁都不值得,真的。生命要掌握在自己的手里,对他到底是一件什么样的事情呢?就是说呃有一个日本的团体哦,那个团体是韩国韩国的啊,对一个日本的粉丝,那么他去看了一个韩国的团体,然后呃他算是追星嘛。嗯他呢也算是花过重金的,就是比如说一些签售会啊呃他也是要。过那种合影的。因为之前我们在节目里也讲过嘛,就是如果能够跟偶像合影,或者是能跟偶像一对一的签售,这件事情是要花多少银子的。之前我们在节目里已经提过了。那么也就是说他是花重金的去做过这些事情的,呃,用郭德纲的话来说,就是他是衣食父母。那么作为衣食父母去看偶像的时候,那么有的时候啊,当然了,也不是说。一定要逼偶像去做什么事情,而是说他呢可能希望偶像能够有犯傻,说所谓的犯傻应该你也知道的,就是我不知道是什么意思。犯傻的意思是比如说我喜欢某个偶像,然后这个偶像给我一情绪价值上的东西,比如说给了我一个wink嗯或者怎么样而这种就是属哎就是一种feed就是给我一个回应,这种就是属于犯傻的范畴。那么那一天呢就。他去看那个偶像,然后那个偶像呢在台上他好像是有一个什么东西,希望能够能够递给那个偶像。嗯,那么么他是一个团体嘛,那团体里面的另外一个人拿到了这个东西。嗯嗯,但是呢就是他想要拿这个东西的人呢他没拿到,哎,那个成员没拿到,其实偶像本人其实这个事情很小我就想很小就很少见的事情,主要是那个偶像,他本人他在就是可能也。演唱会之后还是怎么样,就是有直播说过。对呃,提了一嘴。对,因为其实这个事情呢在也很多现场的粉丝也看到了。嗯,然后这个偶像可能。对,这个事情我我我也不知道他当时是怎么想的啊,对他他可能是无意。嗯,对吧我我我不确定啊,但是他因为他这个说法其实现在来说也是比较有争议的。因为不管怎么样,因为你这个提提的这个提了一嘴这个事情导致的结果。就是那个女生米娜本人她就走选择了一个很极端无端的方式。因为他被大面积的网报,他其实是针对这个事情已经呃一开始。是大家都去纷纷指责他的时候,他已经是道歉过了,好像啊呃而且他并且他选择观战。因为他原来也算是一个小粉头吧他还有一个专门追星的账号啊哦对他有一个自己自己的账号,还有追星账号,让为他都选他的生活账账号也被冲了嘛。是的是也都是被好像那种所谓的开河啊,接底啊,或者也可能有那个哇他的黑料。对,造谣也可能也有,反正他道歉,道歉呢好像也他的道歉也并没有让呃其他人满意。对,没有让别的粉丝对,就没有也有为他说话的,只是骂他的更多吧。我我呃应该是骂他的人的更多,也也是有那种为他为他说话的。但但就是还是被那些骂声给淹没了。对,其实就是我们想要的互动。然后那个互动没达成。对对,然后。可能作为粉丝的话,心里有点不是很愉愉快。对,作为偶像的话,他就他觉得我的体验也不他也觉得体验不好,我我也觉得体验很不好。那我可能呃拿粉丝当朋友了,我就开直播的时候,我就就说出来提了一嘴事情。对但是呢他他估我不确定他知不知道他随口的一嘴会有千军万马的,他的粉丝会替他出征。其实我想去讨伐,但我我在我的视角里,我的视,因为你知道我是不追任何的,我在我的视角里。绝对会呃就你说出来的任何一句话,都会被粉丝,你的那些深深爱着你的那些粉丝们各种解读和对,或者是过度解对他们一定会为你出征的,错错杀100错杀100都不会放过一个的对我我觉得很恐怖的这些事情,循着网线就追过去了,追到米娜本人的账号以及他的追星账号,追星账号。关停了,他自己说他clo了,就是把它关掉了。嗯,就是instagram上面的个人账号。其实他也算是个小网红了,他还有8万粉丝呢,就是在X上面还有几万粉丝,也就是说他加起来全网的话,应该也有二十来万粉丝的。再如果再加上他本来追星的账号的话,那可能就更多。那其实我觉得对于这样子的小网红来讲的话,他也是普通人,对吧?他也是经受不住这样子的。我之前就是可能说的也绝对了一点,就是我之前说过一句话,我就说呃有的时候我们当然不能说。提倡网报,我们当然是反对网报的,但是艺人或者是公众人物,他会比我们多一道防线。确实他的防线在于什么呢?就比如他的助理或者是他的经纪人或者他的公司拿着账号。那么有些东西有些难看的私信难看的评被挡掉,会被挡掉。嗯,所以有的时候呃会有一层保护在公众人物身上,但是像将。自的小网红或者说他其实就是普通人,普通人是没有办法找到直面这些对他就直面这些。可谁谁受得了,你受得了吗?我受不了,我我真的受不了一句,别人骂我我也受不了一句,别人骂我更别说把把你过去,因为你是个人活在这个世上,你绝对绝对所谓的黑料,不会有人是干净的对,不会干净到就说我纯洁白可能谁谁都经不起生挖,是个人都经不起生挖生挖的。所以就是因为可能大家。😊就是道德的这条线有高低。对,但是不触碰法律的红线以上,那不是还是有道德的高低吗?是啊,那特别是女孩子女孩子的道德要求更高,对,对吧?比如说你你不能有这种比如信丑闻呢,对,然后你不能说呃曾经跟网上有跟人兑现呢。对,然后这种东西都不能够有。如果万一有呃然后大家就会过度解读。对,然后就会生挖,然后越挖越越挖越多,越挖越多灶黄摇啊什么之都有。然后我也看见,就是针对这个女生之前就有一些截图嘛?就是说她可能之前在从事一些就是女公关,对女公关就比较擦边的那种对,还有人说她曾经诈骗啊,都有,我我都有看见。嗯,这个就很难,我们我们两个是很难去求证说是否是是真实价真实的。我们很难去考证啊。但我我想说,不管他是怎么样的,与这个事件都没有任何关系啊。即使他是真的就是个女公关也罢,那他也是花钱了的呀,他也为他的偶像花钱了,嗯,对吧?无非就是我花钱了,我就想要一个情情绪价值嘛。对,追星不就是追星图点吗?这个对吗?对,而且如果说他真的做了犯法的事情,法律会制裁他。对对,然后我觉得一码是归一马是的是的,然后私德的问题,那那我们就我我不太管得着,对不对,根本就管不着嘛。是的吧?他他选择哪哪样的生活方式是他本人自己的选择,更何况啊这道信息都不一定是真的对,不一定是真实的。因为任。何事件只要有一方是个女生。他总会跟这种事情沾边的,总会有这样子的消息出来的。说这个女生怎么了,要么对吧?都会有的。是的,所以呃我看了她,因为我也是通过一些手段看到了她的本人的账号嗯。真的是个好漂亮的女生。所以我我昨天看到这个你发给我这个事情,我就觉得好惋惜,这么年轻。他好像今年3月份才刚刚大刚刚毕业,毕业,对太年轻了。我就觉得很很痛心。是的,而且还还让我更痛心的事情就是他下个月他说他下个月要去留学了。嗯。然后我就想说啊,他他其实有光明的未来,光明的未来有很多人生的选择。因为我们是人到中年了。就是你在回想自己年轻的时候,那那段时间就是你什么都没有,你就是有大把的青春。然后你你你想到以后总总会觉得以后的身份会有很多。我以后可以干这个可以干那个那个时候是会有畅。那那个时候对未来的畅想都是觉得自己可能了。对,都觉得有机会。对,然后他就处在这样子的一个就其实是很有希望,很有希望的一个。阶段,但是就因为生命就因为这样子的事件终止了,对,才让我感到特别的惋惜,更更加的嗯惋惜和痛心。其实我我看着他的脸,确实很可爱,也很漂亮。我就想说怎么就下得去口,这样骂他,嗯,就是非常难听的话,就是呃怎么说呢?因为他是各方语言啊,有英文的,有有还有其他韩文的日文。对各式各样的语言。对,就是都在骂他。然后呃网页会自动翻译嘛,翻译成中文让我看到非常难听,就光翻译成中文的时候,我看到那些字眼,我都觉得特太难听了,真的太就怎么怎么说的出口呢?我不敢想象你就是你电脑或者手机那端的你就是那端的人是是到底生活中你你是什么样的人,我。我无法想象,包括还有一个人是在他已经去世了以后对,还在骂还在持续的骂甚至骂依然还是骂的特别难听的话,甚至有粉丝,他们并没有很惋惜,或者是觉得一条生命就这样子消逝了。他们惋惜的反而是怕舆论对偶像本人不利。对啊。我觉得这个这个事情真的很荒谬,就大家现在对生命漠视的程度已经到了这个地步了吗?就我时常觉得这个世界疯了。对,特别是就真的就是这个世界封在我眼前的时候,我都我还依旧觉得太荒谬了,很荒唐和偶像的生命难道真的比你的生命更重要。嗯,对啊,我我不是很我不理我不理解这个事情我说一千到一万。还是因他而起,对对吧?他倒没有名旨,只是他可能说了一些比较有争议的,然后他的粉丝就替他出征了吗?如果他不说,可能这个事情就没了呀。他不说这个事情是铁没有铁没有,就是这么这么这么微小的一件事情。我其实是心痛之余,还是感到很愤慨的。我我也很愤怒啊,真的很愤慨。我就觉得你们把他当成宝那皇帝另外的生命就不重要,不重要了吧,可以用这样恶毒的字眼去。骂别人。😡去网报一个人,他是什么十恶不赦的人吗?对啊。太费解,他也不是私生啊,对他也他也不是私生饭呀,太吓人了,真的非常的恐怖。我觉得是恐怖,极其又荒唐又恐怖。我我都不知道说如果再严重一点的事情,就是比这个什么递东西严重的事情会会会怎么样会造成什么样的后果。我昨天不就跟你讲了,我说还是挺感谢哦对对对对,我太感谢。😊就是有的时候我们不是外网也是有好处,就是有些政策毕竟是双刃剑,我觉得真的因为我们的人口基数太过于庞大了。然后你说这种年轻的就是青少年。嗯,其实现在很多是追星的主力军。对,特别是未成年其实是主力军主力军。然后你看呃如果我们真的可以那么方便的去上一些外网的话,。那大家肯定更希望就是小孩子也好,或者说是作为我们也好,我们可能也也是更希望能在外网有个账号,对吧?是地球村嘛,大家就是可以一起交流。但是正向的是这样的,正向是哎大家一起交流,我也可以交到更多的国外的朋友。对,但是如果我们一旦遇到。不好的、负面的。一些事情的时候,那我们就会遭受全世界人的网报,太吓人了。这个米娜就是遭受了全世界喜欢这个这个偶像的粉丝的网报,太吓人了,真的好恐怖。就是当你看到无数不同语言跟你国家不一样语言的人都来骂你,可能还会有最难听的最难听的话,可能还有地狱黑。对,叫你滚啊,这种这种滚都是件很。很很文明的话了,我觉得就是所以说所有的恶意,他形成了一种怨念,那种念力很差很差的念力,涌向一个普通女神。对。这个其实就是个普通女生,这个对一个人的普通人的精神打击啊非常大非常大。那他还没有那层。我说的那层保护。对,就是直面这些直面这些。骂他的人,我觉得就是abuse了。对,这件事情发生到现在。呃,当然了,那个偶像本人有没有出来声明,我是不知道,但是我看到了一些。哦,粉丝站哪什么的出来为偶像说写了很多很多写了很多,然后非常的冷漠的作词。我。啊,他在里面的就是用用一整段话来说。Mina.的一些黑料如何如何,然后。其实他已经去世了,他们甚至连一句。逝者安息。都都不会讲对他们甚至还在怪他这些舆论风风波,就是他还在怪他说哦会把一些负面带到他们的偶像身上。对,但是这是一条生命哎,一条生命已经。不见了,然后你你们依然还在在乎你们偶像的名声。对。很担心他好像因为这个事情会影响到他的事业。对对,所以我觉得很荒荒诞,就就就是嘛,就是人家只是失去了一条腿,而你。而你可是失去一个舞台耶。嗯,对,这让我真的是我觉得倒反天罡。现在某一些极端粉丝。难道连人的基本都没有了吗?就人性的基...(已截断)
Modern Wisdom
(1)
Q&A: Becoming A Dad, Health Update & Favourite Peptides - #1134
1h 46m
2026/08/08
📝 AI
总结
Hello, friends, Wele back to the show. It is another Q And A episode. And as is tradition, I've asked for questions from Twitter and YouTube community and Instagram. But I also know that you guys have wanted more Q and As with that in mind, if you go to Chris Willex dot com slash stories. You can submit any story, question, live hack problem that you're dealing with, you can even submit a voicemail. and I am going to use that.To speak to guests, I'm actually gonna put some of these stories across to guests. So maybe if you submit some business nightmare that you're going through at the moment, Alex Homoi will react to it. or Jimmy Car can roast your most recent horrible first date. Chriswell X dot com stories, that's always gonna be open. But if you dump them in straight away, maybe you'll get to be on one of the first episodes that we do this with. So that's always on. But given that it's a QAn episode. Let's get into it.😊Bryson rebelbels, many of us are becoming sick of the self improvement space. What's your take on this, dude.I've been feeling this murmuring for a while. I licked my finger and put it in the air, and I felt the winds are changing.I don't know what it's due to, maybe it's an uncertain future AI coming to take everybody's jobs, geopolitical uncertainty, but I know exactly what you mean there is this like self-improvement fatigue that a lot of people are dealing with and also very grind sloppy content of fuck your feelings just work harder a lot of the time that feels like burden homework overwhelm and I think people they just need a little bit of a break, one of the reasons that I'm doing more roundtable episodes,Just hangs with the boys is.I kind of feel like that sneaks some of the self improvement desire in. But in a rapper that doesn't make you feel so much like it's a.A task that you have to complete like another to do list. Sometimes you listen to a upon like that's so amazing, but you feel guilty because you haven't made a anque flashcards or put a ton of notes sound in your Apple notes for it. And yeah, another interesting insight I learned was that in 2020, the biggest financial niches were all slow investment4 a1 K dollar cost averaging S&P. And this year over the last 12 months, the biggest financial niches have all been get rid quick, make money in no time.😊Kalhi Polymarket, like vibe betting your way to a quick million dollars. And I wonder whether this short term missed change is sort of contributing to this.Desire to maybe not think so long termism from a self improvement standpoint. I don't know. But I, I feel it. And I'm trying to create the balance. The show has always been about helping people to understand themselves in the world around them. But also delivering that in a way that's like, oh, I can just sit back and relax. Let me the episode with Alex Peus,2,2 and a half hours about the Odyssey is straight fire. Like you, you don't need to remember anything that he said, you can just sit back and enjoy an amazing retelling of the Odyssey by a guy.With a beautiful voice for two and a half hours and be like, I don't feel like I wasted my time. I do feel like I learned a lot, but I don't feel now like I'm obliged to go and change my entire life in future. I'm going to keep on trying to walk that line. I'm glad that you brought it up. I think a lot of people. like even me, like even I'm feeling this way, right, I'm the tip of the fucking spiritear of this stuff. so.😊I'm going to try and help as best I can.Ynder03 X, How do you know when it's time to break up with your partner, I'm currently in relationship with someone who has a really good work ethic and is constantly busy chasing his dreams, which kind of leaves me feeling lonely 90% of the time. I want to be supportive, but it is hard to be when you are constantly questioning whether your partner still once this. or if he is using busy as an excuse to not spend time with you.哈。Well, I'm sorry that you're going through that I.It kind of doesn't matter whether he still wants this or is using busy because either way, you still feel lonely 90% of the time.I don't think that you should be trying to use your brain to convince your heart and your gut and your body and your emotions that you feel a different way to the way that you feel so much of relationships, especially uncertainty around leaving relationships, is you trying to convince yourself that not feeling good.Is okay. Well, look at all of the reasons why I should be, okay, he's got a great work ethic. He's chasing his dreams, you know.And yet, I feel lonely like this presumably is a guy that's reliable and and hardworking and stuff. All of these things are great. But you feel lonely 90% of the time I.I don't know what else to say other than that's not what a happy relationship is supposed to feel like.You're not supposed to feel lonely 90% of the time, and.You can try and bet on a future where his ambition, fuel runs out, but you're basically in love with the potentia...(已截断)
All-In with Chamath, Jason, Sacks & Friedberg
(1)
Google's AI Brain Drain, SpaceX's Huge Quarter, Airtable's 90% Collapse, US Data Fuels China AI
1h 15m
2026/08/08
📝 AI
总结
Alright, everybody, welcome back to your favorite podcast. It's the all in podcast. It's the summer. It's August 6, having a hard time getting a quorum here on the podcast. But David Freeedberg is here. David Freeberg is back ourultan science. How you doing, brother, Great to be with you. It's great to be with you and.😊Everybody loves.Brad Gerersner is here. Hes your Bruce Wayne. If markets are your game, he brings that note to your pay day. Yeah his glasses that discount. and'll get you one of those fancy Trump accounts. All right, welcome back to the program. Brad, I love it. I love it. You're bringing the rhymes back I bring a little intro back. We've been Chm is on the road right now. Chath is on the road, but we will get a field report from Chm and I call Daniel, somehow Sas is gonna to be here, but.Younow how he is. He's always late because, you know, you get a phone call from very important people, But he will break in at some point. Oh, wait, I see in the text area. Oh, Oh, there is. He really does. You made it. How do you like my beautiful summer Jo. It's incredible. It fits perfectly. You look warm. I don't know how yourath does this.😊🎼あははあは。Let your winner slide.🎼Rain man, David out.We open source it to the fans, and theyve just gone crazy.🎼Lo me queen.Well, here's the report. everybody. As everybody knows, Chm is on the road. he Oh, here he is. He this is a photo Ss. He went, He went to check his data center progress. I think that's in Colorado or Nevada, where he's building a data center I that's onne. Oh, it's on Done. Yes, Junene 4. Oh yes, here is admiring himself. Oh, look, here's. You know when a meme has reached its peak when your wife starts sununking on you, there it is. And here we are.This was at the Christmas party, I think. Oh Brad, you were on CNBC Andrew there I wasn't sure if it was my Twitter feed that was just selecting into it, but it not only hit everyone right This is a viral thing. This has hit everything.less we got a lot to get you enough with the shenanigans and small talk. Google had two major shakeups to its AI staff on Wednesday. Des Haabs has moved to chair of Deep mindd and chief scientists at Google reports described this as Demis stepping down or being.😊Kicked upstairs we'll get into that, but Google framed it as of promotions and says he was stepping up Here's exxiososis quote explaining the shakeup quote Google's Gemini 3.5 pro is months behind with some company sources telling Exxios that it's in part due to low morale, interesting several top researchers including Gemini's co lead have left the firm for competing AI Labs, Che Dean plus three other AI superstars we leaving Google to start a company called Discovery loop Dean is a legend Freeberg and they think you.Work with him at Google, one of the world's great AI engineers. He was employee number 30, joined in 1999 and his worked there for what I understand continuously for 27 years, Discovery loop's going be focused on deep scientific breakthroughs in AI. Google showss down 4% on the news of Dean leaving. So 200 billion in lost market cap. If you want to correlate those two things. Freeberg. This is your alma mater. What are your thoughts here, I this creative destruction, Maybe these people aren't delivering and.They wanted fresh flood or is this just the siren call of doing a startup in an age of unlimited capital for AI and unlimited opportunity, just being too much for the OGs at Google to not take advantage of Maybe it's the third bucket, which is if you're the board and the management, you're having a debate about how to best deploy capital Google has made a commitment to deploy $200 billion in Capex this year.In AI infrastructure, data center build out because of the Capex and accelerated depreciation.Making an investment in AI compute in the US right now is hugely tax advantaged. And because of the extreme demand for compute.It's a pretty obvious kind of ROicC model return on invested capital. So if you make this sort of an investment, you have significant demand for that compute infrastructure. You're very good at running the compute infrastructure. That capital can deliver massive profit returns for you with very high confidence in some forecasted period, building the most advanced frontier lab driven model. also takes tens of billions of dollars of capital and the.Question reallyally is, can you deliver the profits from the model.And in a world where open source is becoming so good and open weights models are catching up so quickly. and all the frontier labs are catching up to each other so quickly. Does it really make as much sense to deploy tens of billions of dollars against building a model. And I think that the scientists that were're seeing transition out or the scientists that have been at the core of model development of making these frontier models? And they were certainly first out the gate.😊You could look at some of the early interviews with Jeff Dean from a couple of years ago where t...(已截断)
Lightcone Podcast
(1)
He Built the World's #1 Open-Source Coding Agent
N/A
2026/08/08
📝 AI
总结
🎼Most people in the world still haven't experienced the magic of a coding agent. This is just an unprecedented market. Like the market for intelligence has not existed before. Everybody should be thinking a positive sum, for the high mentality. It's huge in developing countries like Indonesia is 4% of your traffic. Brazil is 5% of your traffic, places like Vietnam, you know, like places where a $200 a month plot could subscription is like very expensive. You really know you have product market fit when like enterprises are bugging you.Me sign the security agreement so they can get to your product. like, please do this, because I don't know who you guys are, but a bunch of us are using this. We want, you know, everybody in the world to have that aha moment with a coding agent.Welcome back to another episode of the Like home at Gary's Out traveling today and we'll be back next episode. Our guest today is J V, founder and CEO of Opencode, an open source alternative to Claude code that works with any model you want. Open code has been growing at an astounding rate this year. Then now at 4.6 million weekly active users, which is actually pretty close to codex's numbers. Today, we're going to talk about what's driving their wild growth. and also Jay's winding road to get here since the company went through YC back in 2021. Jay, thanks.So much for being here. Thank you. Thank you for having me. Why don't we just start with kind of like the, the crazy scale you guys are at. Maybe tell us about any stat you can share with us. Yeah, yeah, I mean, you mentioned the, the weekly actives are monthly actives. I think we ended June with around 13 million or so.😊And that's around a 20 x increase since the beginning of the year.We also recently started processing around 7 trillion tokens per day for context, open router does a total of around 6 trillion. I think, again, sort of beginning of the year, we're probably at around 300 billion or so.And then in terms of sort of our revenue off of our subscription product. And if you were to paper token with our inference.And you took, let's say, June's data and extrapo later for the year. That would be around 31 to 33 million or so. And if you took last week's data, an extrapolated for the year, that's around 38 to almost 40 million. And that's the inference part of our business So the way we sort of make money. We had launched that.Call it end of September, early October last year. So about 8 months to getting to around 40 million or so. I just to sort of round out the numbers.Are subscribers. So these are people that that pay for a monthly subscription with, with open code.We launched that product.Early March at the end of February. And that's grown to around 160000.Monthly subscribers. Yeah, and around that, that accounts are about 18 million of sort of the annualized revenue. I sort of tweeted from I think I think he is Tpo. I think he's a lead engineer, at least one of the main engineers on codex saying that something like, I guess5% of all codex subscribers choose open code as like the main harness to actually use the, the underlying API. Yeah, yeah, so Codex officially, you know, supported..Open code. what that basically means is that you can use code access subscription in open code. And yeah, and, and a bunch of their users use open code directly to take advantage of their subscription plan. And I think they did that right after kind of some of the some of the the back and forth you had, maybe with anthropic and crawl code. Like tell us about what happened in how that seemed like it really fueled your growth It was an inflection point for you guys. Again, to sort of contextualize this. We were at around 650000 monthly active users to begin the year. And.First week of January, we start to sort of hear some rumblings around anthropic, trying to clamp down on.People using open code, but using cloud code subscriptions on there. And this was a very sort of common way to kind of use the cloud code subscription. And the way I think that that sort of blocked it or they were, were trying to block it was if the system prompt mentioned literally the word open code, they would sort of reject the request. You know, from our perspective, that sort of makes sense. You know,, they're sort of subsidizing usage, that's kind of what they want to do. But of course, a lot of users weren't happy. And when they had kind of done that.What it inadvertently did was it put open code and cloud code on the same sort of pedestal. It like equated the two products in some ways and.And even for the people that weren't using open code at the time, you know, they they sort of took notice of the fact that that, you know, cloud code is is taking that kind of an action against against open code. You got new users because people heard about you for the first time with they' like, oh cloud codes like banning this thing or like clamping down this thing. Yeah, or that it it is like worth looking into that, you kno...(已截断)
Dwarkesh Podcast
(1)
8 Predictions for the Era of Continual Learning
8m
2026/08/08
📝 AI
总结
🎼So it explained elsewhere why I think actual continual learning is needed. I don't think you can have AI that perform whole jobs as competently as humans if they are forced to just write marked on files from session to session. just to give an illustrative example. Imagine if this is the way the students learned to play the saxophone. So you have one student. he's never played the saxophone before. He goes into the music hall. He tries to play it. Of course, it his first time. So he fails. And he writes down a bunch of notes about what went wrong. And there's a next student who's waiting outside the music hall. He comes in. He reads all these notes. He's also never played. So of course.Messses up. And he continues to add on to these notes. And you have an infinity of students who are outside the music hall who keep writing notes to the next person. I don't think there's any sequence of text they could write to each other that would allow the subsequent student to just nail the saxophone from the first try. at some point, do you actually have to accumulate the relevant experience into your brain. I think the same thing will be true for a lot of skills that we want AIs to actually accumulate from all the different workplaces in which they're deployed. Okay, so what changes once we have actual.😊al learning1, I think that a lot of proposals that have been put forward about regulating AI assume that you train a model and then you deploy. And therefore, if you run a bunch of checks on the model before it is deployed, we can make sure that it's not gonna to aid in cyber attacks or do something crazy. I don't think this assumption necessarily makes sense in the future. And this is one of the many reasons I'm actually kind of worried about locking in some kind of safety, regulatory regime right now, because we don't know what kind of technology we're going be dealing with, even within any year.Let alone within five years or 10 years. What if the model is improving every single day based on the millions of sessions of work it does in that day. If that happens, we could potentially be locking in an archaic and potentially counterproductive approach to dealing with the threats from AI. to the extent the government wants some way to do some kind of safety valuation on model providers, I think it would make more sense to do monthly or quarterly risk inspections, rather than trying to single out some special moment that occurs after trading is done. But before deployment begins, because that will not.Be a meaningfully distinct category in the future,2, how the labs do technical alignment would probably totally need to change Right now, a lot of research is focused on the question of how we make sure that a frozen set of weights behaves well during deployment. But I'm not aware of much research on the question of how we make it so that even with constant weight updates, the AI system never falls created jail breaks or changes into a deceptive or evil persona. And if AI is are consolidating learnings between users as well. How do we prevent.Users from injecting backdo or some kind of malicious inclination into the baseball. In some sense, this is actually kind of what the human alignment problem is, right, Humans improve in a selfdirected way. If you have kids, I don't have kids, but I imagine this is what happens. you have kids, They go out, They learn new things. Sometimes they go crazy. They get one shoted by crazy ideologies, They take the wrong drug. They become super weird. But you hope that you've given them enough common sense. and basic values. that they improve as people in a selfdirected way.😊Without ending up with some super weird beliefs or some misanthropic ideas,3, the diversity of AI minds will increase right now, there are less than five prominent AI minds, by which I mean, the base models, which are served to millions or hundreds of millions or billions of users at once. And they're all quite similar to each other, by the way, because they've also been all trained on roughly the same data. But if AIs are learning from experience and that experience is different between not only different AI companies, but also between different instances of the same AI.Model, we could actually see a lot of diversity come out the other end in this world. And this would be, I think, a net good outcome. I think one of the things to worry about in the future is just having this monolithic singleton that's quite boring. a world where we have conal learning would hopefully be more interesting, then the mode collapse of different models we see in the world right now. for, when deployment becomes part of training, the returns to being ahead in the AI race accelerate because if you have the best model and more people are using your.IFor more complicated and useful work. And as a result, they're giving lots of feedback that can integrate beyond the session window. Then your model will become even smarter.5, if the model learns main...(已截断)
Y Combinator Startup Podcast
(1)
How To Design In The Agent Era
56m
2026/08/08
📝 AI
总结
🎼The agents can speed you up. They can save you a lot of work, and they can do tasks like translation or resizing for you. But we're big believers in the human element of design. and how important that is. And so we're trying to create a tool that lets humans work really fast with agents. It's probably very tempting to use whatever cloud design spits out. I don't think it helps you stand out from the crowd. And so I think design is this great differentiator. And if you look at every great company of the last 1020 years. Basically, all of them have exceptional design. And so I think if you want to be one of those really great companies, you have to put the care in on design. And and make sure that you stand.😊あ?🎼Today, I'm excited to welcome Stephen Haney, the founder of paper, an AI native design tool that's been taking the design world by storm. We're already using paper a ton here at YC for all of the design work that we're doing internally. And I'm excited for him to show off how the product works to everybody out there. So, Stephen, thank you so much for joining. Yeah, I'm honored to be here. I watch the show a lot. So it's it's really nice to be, to be on. Thank you for having me. Maybe to start tell people what is paper. Yeah, well paper.Is a design tool. It's an agent native design tool. And what that means is we use HTML and CsS as the actual rendering engine. And that makes it very easy for agents to use because they understand HTML and CSS really well. And historically, a lot of design tools have custom rendering engines. And those are a lot harder for agents to use. And so you'll see higher token spend, slower times, lower accuracy. So what we're building with papers is, basically, it's a group of people is about 12 of us now, all designers and engineers, we were basically started this.Of love for design. And now we're just trying to build the best design tool possible for this next era. Yeah, it's crazy to think you're only 12 people because I hear about you all the time from everybody who's using you. and especially from so many founders that are just picking the design tool that they want to build and live in all the time and doing that from the ground floor being AI native themselves, They're looking for the AI native tool, that is going to work best So maybe talk about, what was your inspiration for building paper. And why were some of the existing tools that are out there, Not good enough for you.Re was born from a love of design. As I said, this was, I started as a solo founder. And it was really the side project that just kind of kept going. I had months where I was like, is anyone gonna care about paper and that it was like, oh, yeah, people do care. You know, this is great. It's really taking off. It was the thing I was gonna do anyway. It was very much the thing I was gonna do anyway, which I really believe starting a startup. You should do that. You should do the thing you're gonna to do anyway. You should do the thing where you have founder market fit. And for me, This is where I had the strongest founder market fit. because I love building tools for creative professionals, It's something I'm gonna wake up every day.And like I'm not gonna have a bad day building creative tooling So that came first even before the market or competitors or any of that kind of thing and then looking around it was like, well gosh, I mean know we're big fans of Figma you almost can't talk about design without saying Figma these days they've done a lot for the design industry they have a huge company now and they're doing very well with enterprise companies and we're thinking well maybe there's time for a new tool to focus on the designer and just really spend a lot of time there's so much change with AI the agents can we build something that's agent native building on each team.CS SS and just build this, you know, next gen tool. And so that's kind of what we're going after. And so far so good. Yeah, and what does that mean to be an AI native agent native design tool, It's really about the fundamental technology. So using Ht and CS as the rendering engine. We actually did it to help with designer developer handoff. We were like, hey, if the designer can have this constraint free system where it feels like a figma or a sketch, But by the way, it's Ht, they don't have to know that. That's great. The handoff will be better. It turns out if you build a great handoff.Humans, you've also built a great handoff for agents because agents understand HTML and CSS and it's in their training data. And so when you ask an agent to work with a paper file, it's literally just reading the CSS app and agents know exactly what to do with it. You try to basically extend the agent with a visual interface is sometimes paper is a design tool, but it's also this visual interface for agents We see sales, engineers, all kinds of people are using it because prompting isn't always the best input mechanism for agents. Sometimes you want...(已截断)
The Joe Rogan Experience
(1)
#2537 - David Sinclair
2h 30m
2026/08/08
📝 AI
总结
🎼The Joe Rogan experience during my day, Joe Rogan podcast by night day.Mr. Sinclair, good to see you again.😊🎼Dorctor Sinclair, doctor, I'm kidding. better than David. Yeah, Do you ever introduce yourself as doctor, Never. Yeah, it's a kind of a. It's a weird thing because you're allowed to.I'm actually a so David from Australia. but are you a Sir. Yeah, but it's kind of a joke. I don't really. What does that How that work out, Did you get united or clean representative gave me some Australian version of a knighthood.But I get to put A O at the end of my name. A O, Yeah, Order of Australia. Oh, But it actually makes no difference to my life. So it's just an extra couple of letters. It's kind of cool. Well, I man, I'm honored. for sure. But it it's just a joke. you don't name. What's kind of anybody. Sir, you meet a guy at the gas stage. How you doing, Sir.😊Is that like, wow, how does he know, right? You don't say how you' doing King.Yeah, right, well, I'm David. I'm Australian. So in Australia, if you have any ego, you get beaten up in the playgrounds. That is true. Australia is the land of the tall poppy syndrome. They don't, They don't fuck around. They don't like you getting a big head in Australia. That That is true. I like to go back there, though, because it's any level playing field. Nobody cares who you are. I love the people there. So Australian people are some of my favorite people. They're awesome. There's, they're very friendly and fun. It's great place to be.😊It's too far to get there. Such a hall. Oh, no, if you just, you know, whiskey and an amb and and you're there without. Is that what you do, I used to. I've gone clean. I don't drink much at all anymore. Last time it is mean how often it is drink much at all. sell drink right now. No, it would be a bad look.I've got the coffee that you gave me. which Yeah, this is a laird superfood coffee. So Laird Hamilton his it's black rifle coffee mixed with his superfood stuff, which is this one's turmeric. So it's turmeric. and I think it has coconut milk in it and some other jazz. It's delicious. anti inflammatory. Yes, they've done, which I know you're all about. yes, as a, a profit of longevity. Yes, is that, is that another title of you. Yes, you can.😊Call me, Sir, David, the prophet. How about that.Sir David, the prophet, Doctor, Sir David, the prophet of longevity. that do. So there's a lot. Let's just jump right into it because there is a lot going on in the longevity space. Thank you having me back on it. My pleasure. And by the way, I also want to thank you for. I think this is both time on. which is amazing. Thank you. Congrats on all your success. It's thank you. Wo to see.😊Also, the book that I wrote, which you first had me on to promote, has done so well. And it's really because of you. And the book lifespan became a besteller, not just that, but it's now been read millions of times around the world. It's in 4. That's phenomenal. And it's my pleasure. And that's not a promo. as you know, you don't make any money from books, really, But I'm just honored that it was read by so many people around the world because it wasn't an easy read. It was high high tech science. Well, people are very interested, especially nowadays. It's I think.Never, I can't recall a time in my life where optimizing your health and longevity has been more of a subject that people are interested in. Well, I think that's largely because not to keep blowing smoking up your rear, but largely because of what you did in the book. So the book went global. and it's, it was right in 2019 in the pandemic hit.And I think most people saw their own mortality at that point.Most people under the age of about 70 didn't even think about dying. but then pandemics, like shit, I could die.And I saw the book take off. You launched it. But yeah, since then, a lot of people have jumped on the bandwagon. and now longevities. I hear this everywhere I go. It's everything. It's everywhere. People are constantly talking about different modalities. How often do you cold plunge, Are do you saw it. What about the red light, What do you think, I this legit, Is that legit, like this, everybody's optimizing everything constantly all the time. But the reason why it's caught on is because if you do follow some of the protocols, you will feel better. And so that's.The ultimate goal, like longevity is wonderful, but feeling better in the moment currently, like living your life should be primary. longevity, I think, is if you're optimizing your life and you feel better, you will live longer. It's like this, it's all part of it. would follow the longevity principles and the ideas behind it. You, you certainly will live better right now. You will. And it's not even that hard. There are Ia things,5 basic things you can do or avoid and you can live another 50.Use, it's really not that hard. And that's what I'm excited about telling millions of people today is. And maybe we can go over the kind of things that...(已截断)