@MinLiBuilds 可能是Ollama不是Mac最优解？AI推荐我vMLX 32G M4跑Gemma 4 e4b也不快 反应都要40秒左右
中文
旋风小鸡腿
785 posts
旋风小鸡腿
@speedydrumstick
提高认知ing | 一只叫小鸡腿的猫的散户私人投资笔记和心得｜to learn how to ask the right questions
Beigetreten Ağustos 2009
316 Folgt46 Follower
Anthropic的工程师Kevin的一个分享。
价值$1000 刀，关于agent memory的workshop。
这将是 AI memory的主流，提前一年了解它，成为记忆专家。
非常值得看的一个短视频，以下是视频中的实操：
03:31 证明 AI 多个对话间没记忆（对照实验）
06:20 介绍 memory store，可以作为 AI 的记忆外挂
08:44 将 Memory Store 挂载给会话，实现跨对话记忆
15:36 执行 Dream，让 AI 在后台整理记忆并扩写简陋笔记
19:45 展示 dream，记忆变成了结构化
看完这个视频，接着阅读下面的这篇文章。
你将充分了解与 AI agent 一问一答背后，记忆的原理，以及最强省 token 的心法。
实践哥MinLi@MinLiBuilds
中文
Elon 永远是我心里最酷的男生，没有之一
AYi@AYi_AInotes
卧槽！@SpaceX 这个构想直接给我看傻了🤯 在地球和月球之间拉一张光网， 近地轨道的Starlink，对接月球轨道的中继星座， 以后月球上的人刷X、打视频电话，和地球没有任何延迟， 我尼玛，这哪是商业公司干的事啊， 简直就是在给人类铺星际互联网啊， @elonmusk I want to say， 最近发生的事，对我来说最震撼的不是是火箭和这个月球星链， 是你总在所有人还在争论能不能去月球的时候，已经在铺去了之后怎么生活的路， 谢谢你总在提醒我们，人类的边界从来不是地球， 等以后有人在月球上刷到这条推的时候，我想他一定会记得今天这个视频的，Always respect，地表最帅最强的男人🫡
中文
1️⃣我接棒负责家里的一日三餐之后，我发现：
- 如果三餐，有鱼，有肉(牛肉)，有海鲜。
- 很贵啊。
2️⃣不是一般家庭能承担的。
中文
AI 时代真的来了，我老婆快过生日了，跟我不要香奈儿不要 LV 了，跟我要了一台苹果 M5 26➕521G 硬盘的 MacBook air
中文
旋风小鸡腿 retweetet
New for financial services: ready-to-run Claude agent templates for building pitches, conducting valuation reviews, closing the books at month-end, and more.
Install them as plugins in Cowork and Claude Code, or use our cookbooks to run them in production as Managed Agents.
English
OpenAI总裁当庭认罪！自曝零元购300亿，马斯克这回真要赢了？
太炸裂了！刚刚，OpenAI总裁Brockman当庭承认：自己投入0美元，持有OpenAI营利部门300亿美元股份（马斯克捐了3800万，得到的是0）。更炸的是，Brockman和奥特曼都悄悄持有Cerebras个人股份。Gary Marcus直言，这是马斯克最接近赢的一次。
就在刚刚，OpenAI总裁Greg Brockman「认罪」了？
他当庭承认，自己从未投入一分钱，却套出了价值300亿美元的股权。
这个消息，不仅惊呆了法庭上所有人，也让所有网友震惊。
听到这个劲爆消息，纽约大学学者马库斯判断：我认为马斯克第一次真的有机会赢了。
庭审现场，马斯克的代理律师神情自若，他拿着Brockman自己的日记和邮件，优雅地进行着一场「现场活剐」。
最令全场震惊的一幕发生了。
当律师问及Brockman在OpenAI营利性实体的持股时，对话如下。
问：「你在这一营利性公司中拥有所有权权益，对吗？」 Brockman： 「是的，准确。」 问：「而为了获得这些权益，你投入的现金是0美元。对吗？」Brockman（迟疑后）： 「也是准确的。」 问： 「你在这家营利性实体的股权，以今天的估值计算，超过了200亿美元，对吗？」 Brockman：「是的。」 问： 「事实上，它可能更接近300亿美元。对吗？」 Brockman： 「我想这可能是真的。是的。」
这个数字在法庭内回荡时，旁听席上响起了一阵细微的骚动。
要知道，马斯克作为OpenAI最早的资助者，先后捐赠了超过3800万美元现金，提供了早期的办公场所，甚至亲手挖来了顶尖人才。
但在今天的OpenAI里，马斯克的个人占股是零。
不仅如此，Brockman还不得不承认了一个尴尬的事实：他曾在早期筹款中利用马斯克的名字来背书，甚至曾口头承诺捐款10万美元，但实际上，这笔钱他从未兑现。
这正是马斯克指控的核心：不当得利。
按照加利福尼亚州的慈善信托法，非营利组织的受托人应该领取薪水，而不是瓜分慈善资产。
马斯克的逻辑很简单：我捐款是为了做一个造福人类的公器，结果你们把公器偷偷拆了，把里面的零件装进了自己的口袋，还给自己贴上了300亿美元的标签。
Cerebras：一笔200亿的自我交易
如果说300亿股份是这场庭审的第一颗炸弹，Cerebras就是第二颗。庭审中披露的关于Cerebras公司的关联交易，直接触及了法律红线。
马斯克的律师翻出了一份陈年旧账。
2017年，Brockman在担任OpenAI受托人期间，私下购买了AI芯片初创公司Cerebras的股份。与此同时，奥特曼也对该公司进行了个人投资。
然而，在接下来的时间里，Brockman开始疯狂在OpenAI内部游说，促使OpenAI与Cerebras达成交易。
具体时间线如下。
2025年12月： OpenAI签署了一份向Cerebras支付100亿美元的订单，并额外提供了10亿美元的贷款。
2026年2月： 凭借OpenAI的巨额订单，Cerebras的估值从80亿美元飙升至230亿美元，翻了近三倍。
2026年4月： OpenAI将订单追加到了200亿美元。
现在： Cerebras正式提交IPO申请，估值冲向266亿美元。
法庭上的对话如下。
问：当你在讨论OpenAI和Cerebras之间的金融交易时，你实际上是Cerebras的股东，对吗？Brockman：「讨论期间和作为Cerebras投资者之间存在一些重叠。是的。」问：你能指出一封告知马斯克你持有Cerebras股份的邮件吗？在你推动OpenAI和Cerebras做交易的同时？Brockman：「我不认为存在这样的邮件。」问：聊天记录呢？Brockman：「没有。」问：短信呢？Brockman：「没有。」问：然而如果OpenAI和Cerebras之间有交易，你个人会从中获利Brockman：「我想是的，但这不是我当时考虑的事情。」
加州慈善信托法对此有个专门的名字：自我交易。这种自我交易在法律上是极其致命的。
作为非营利组织的负责人，他们利用慈善资金去扶持自己个人投资的公司，从而实现个人财富的指数级增长。
这已经不仅仅是「背离初衷」的问题，而是涉嫌严重的职业道德违规和利益冲突。
这场审判，本质上是两种硅谷价值观的终极对撞。
一种是马斯克代表的「旧约理想主义」： 承诺就是承诺，契约不可违背。
一种是奥特曼代表的「实用扩张主义」： 技术迭代太快，生存才是第一位。为了实现AGI，任何法律结构的微调和利益分配的改变都是合理的。
而加州法律可能更倾向于前者。
在加州，慈善资产受严格保护。如果你建立了一个慈善机构，然后决定把它变成私人公司，你必须经过极其复杂的评估，并将资产价值全额回馈给公众。
如果马斯克赢了，会发生什么？
首先，OpenAI可能被迫开源——这是马斯克一直以来的诉求。
第二，微软的投资可能面临风险。马斯克要求撤销与微软的独家授权协议。如果法庭支持这一请求，OpenAI的估值将瞬间坍塌。
第三，OpenAI营利性部门的利润可能被强行划拨回非营利母体，Brockman那300亿美元的「零成本股权」可能化为泡影。
最后，这一判决将成为判例，警告所有AI初创公司：你不能打着慈善的幌子融资，再打着商业的幌子收割。
但如果OpenAI赢了，那意味着硅谷的「野蛮生长」逻辑再次获得全胜——只要你能做出改变世界的技术，那所有最初的诺言，都可以在成功的金光下被掩盖。
OpenAI CEO Pleads Guilty in Court! Admits to Buying $30 Billion Worth of Shares for Zero Investment – Is Musk Really Going to Win This Time?
This is explosive! Just now, OpenAI CEO Greg Brockman admitted in court that he invested zero dollars to acquire $30 billion worth of shares in OpenAI's for-profit division (Musk donated $38 million, receiving nothing in return). Even more shocking is that Brockman and Altman both secretly hold personal shares in Cerebras. Gary Marcus stated that this is Musk's closest chance to win.
Just now, OpenAI CEO Greg Brockman "pleaded guilty"?
He admitted in court that he never invested a single penny, yet acquired $30 billion worth of equity.
This news not only stunned everyone in the courtroom but also shocked netizens.
Upon hearing this bombshell news, NYU scholar Marcus judged: "I think Musk has finally had a real chance to win this time."
In court, Musk's lawyer remained composed, elegantly conducting a "live dissection" of Brockman's diaries and emails.
The most shocking scene unfolded.
When the lawyer questioned Brockman about his stake in OpenAI's for-profit entities, the dialogue went as follows:
Q: "You own an equity interest in this for-profit company, right?" Brockman: "Yes, exactly." Q: "And to acquire these interests, you invested $0 in cash. Right?" Brockman (after hesitation): "Also, exactly." Q: "Your equity in this for-profit entity, valued at over $20 billion today, is that right?" Brockman: "Yes." Q: "In fact, it's probably closer to $30 billion. Right?" Brockman: "I think that's probably true. Yes."
As this figure echoed in the courtroom, a subtle commotion arose in the audience.
It's important to understand that Musk, as one of OpenAI's earliest funders, donated over $38 million in cash, provided early office space, and even personally recruited top talent.
However, in OpenAI today, Musk's personal equity stake is zero.
Furthermore, Brockman had to admit an embarrassing fact: he used Musk's name to endorse the company during early fundraising, even verbally promising to donate $100,000, which he never actually did.
This is the core of Musk's accusation: unjust enrichment.
Under California's charitable trust law, trustees of non-profit organizations should receive salaries, not divide charitable assets.
Musk's logic is simple: I donated to create a public instrument that benefits humanity, but you secretly dismantled the instrument, pocketed its parts, and labeled yourselves with a $30 billion tag.
Cerebras: A $20 Billion Self-Deal
If the $30 billion stake was the first bombshell in this trial, Cerebras was the second. The related-party transactions involving Cerebras disclosed during the trial directly crossed legal red lines.
Musk's lawyers unearthed an old case.
In 2017, while serving as a trustee for OpenAI, Brockman privately purchased shares in the AI chip startup Cerebras. At the same time, Altman also made a personal investment in the company.
However, in the following period, Brockman began aggressively lobbying within OpenAI, persuading OpenAI to reach a deal with Cerebras.
The specific timeline is as follows:
December 2025: OpenAI signed a $10 billion order with Cerebras and provided an additional $1 billion loan.
February 2026: Thanks to OpenAI's massive order, Cerebras' valuation soared from $8 billion to $23 billion, nearly tripling.
April 2026: OpenAI increased the order to $20 billion.
Now: Cerebras has officially filed for an IPO, with its valuation approaching $26.6 billion.
The courtroom dialogue is as follows:
Q: When you were discussing the financial transaction between OpenAI and Cerebras, you were actually a Cerebras shareholder, right? Brockman: "There was some overlap between the discussions and being a Cerebras investor. Yes." Q: Can you point to an email informing Musk that you held Cerebras shares while you were pushing for the OpenAI and Cerebras deal? Brockman: "I don't think there were any such emails." Q: What about chat logs? Brockman: "No." Q: What about text messages? Brockman: "No." Q: However, if there was a transaction between OpenAI and Cerebras, you personally profited from it. Brockman: "I think so, but that wasn't what I was thinking at the time."
California charitable trust law has a specific name for this: self-dealing. This kind of self-dealing is extremely dangerous legally.
As heads of nonprofit organizations, they use charitable funds to support companies they personally invest in, thereby achieving exponential growth in their personal wealth.
This is no longer just a matter of "deviating from the original intention," but involves serious ethical violations and conflicts of interest.
This trial is essentially an ultimate clash between two Silicon Valley values.
One is the "Old Testament idealism" represented by Musk: a promise is a promise, and contracts cannot be broken.
The other is the "pragmatic expansionism" represented by Altman: technology iterates too quickly, survival comes first. To achieve AGI, any minor adjustments to the legal structure and changes in the distribution of benefits are justified.
California law may lean towards the former.
In California, charitable assets are strictly protected. If you establish a charity and then decide to turn it into a private company, you must undergo extremely complex evaluations and return the full value of the assets to the public.
What would happen if Musk wins?
First, OpenAI might be forced to open source—a demand Musk has consistently made.
Second, Microsoft's investment could be at risk. Musk is demanding the cancellation of the exclusive licensing agreement with Microsoft. If the court supports this request, OpenAI's valuation will collapse instantly.
Third, OpenAI's profits from its for-profit divisions may be forcibly reassigned to its non-profit parent company, potentially jeopardizing Brockman's $30 billion "zero-cost equity."
Finally, this ruling will serve as a precedent, warning all AI startups: you cannot raise funds under the guise of charity and then reap profits under the guise of business.
However, if OpenAI wins, it means that Silicon Valley's "wild growth" logic has once again triumphed—as long as you can create world-changing technology, all initial promises can be overshadowed by the glitter of success.
@elonmusk @gdb #OpenAI
中文