Liam Phoenix Wayne

捡了块漂亮石头，ai说这是一块黄蜡

Kevin warsh长得很像我大伯

Gemini3.5flash今天突然就好用起来了，谷歌这是浪子回头了吗，今天准确性大大提高了，手里的$googl又拿得住了

@MaxForAI @OfficialLoganK @khushbooverma 说破天也不好用，废话一大堆，干活真不行

在Gemini被骂上风口浪尖的时候，Gemini开发者关系负责人@OfficialLoganK接受了@khushbooverma的一次播客访谈。 他在里面对于很多尖锐的批评做出了正面回应，我大概看了下，主要是以下几点： 第一，Gemini3.5Flash贵。 Logan的解释是：不要只拿它和3.1Flash比，它其实是按更强模型的能力段来定价的。 对成本敏感的场景，Google会推3.1Flash Light这类更便宜的模型。 换句话说，Google在把Flash从「便宜快模型」重新定位成「高智能、高速度、偏Agent和代码场景的主力模型」。 第二，3.5Flash的重点场景已经从聊天转向Coding和Agent。 里面有一句很关键：过去Flash主要服务聊天，现在用户越来越多是在让Agent替自己干活，而这些Agent背后大量在写代码、调用工具。 所以Google要求3.5Flash必须很强地适配这些场景。 第三，GeminiCLI被并入Antigravity，不完全是「Google讨厌开源」，而是他们想把SDK、CLI、Agent manager、编辑器、API里的Antigravity能力统一到一个入口里。 这个解释挺官方，但也暴露出一个问题：Google开发者产品太碎了，过去自己也知道大家很困惑，现在开始收口。 第四，Antigravity的方向不只是写代码工具，而是往「开发者桌面工作台」走。 它会从代码编辑器、Agent manager，逐渐扩展到软件开发相邻的任务，甚至可能变成一个更宽的桌面Agent入口。 第五，最有价值的一段是关于Agent可靠性。 Logan承认Agent时代有一个矛盾：开发者想要速度，但真正上生产时又需要确定性、可观测性、trace和测试。 我个人最喜欢那句「you can outsource intelligence, you can’t outsource understanding」 意思是你可以把智能外包给AI，但不能把理解也外包掉。 我个人还是挺佩服这样的开发者运营。 因为之前大V@theo发视频批评的时候，DeepMind的人也都是在评论区做正面回应。 也没有说辱骂，也没避而不谈。 而是真的就是非常直接的去回应，非常的真诚。

如无必要，勿增实体。常规多伪，本源为神。

@wallstengine These analysts definitely haven't used Gemini 3.5; if they had, they probably wouldn't have given it such a high rating.

BofA Reiterates $GOOGL at Buy, PT $430 Analyst comments: "At I/O 2026, Alphabet introduced a broad range of new AI products, underscoring accelerating velocity of AI innovation. While in prior years Google was playing catch-up with AI technology, this year’s product announcements leveraged Google’s strong Gemini model capabilities and demonstrated increasing leadership in shaping consumer AI experiences. The event, along with strong search traffic data, with AI mode usage doubling quarterly and having 1B users, reinforced our view that Alphabet remains well positioned to lead the next phase of consumer AI adoption, supported by leading frontier capabilities, a differentiated full-stack platform, and unmatched global distribution. Maintain Buy. Key highlights at I/O: 1) Showcased progress in the frontier AI roadmap with the launch of Gemini 3.5 Flash and Omni; 2) Integrated Gemini 3.5 Flash in Search and introduced a redesigned intelligent search experience, further establishing search as an AI starting point; 3) Introduced Gemini Spark, a persistent personal AI agent, and launched Antigravity 2.0, an upgraded agent-first coding platform; 4) Launched Universal Cart, a persistent cross-merchant shopping cart that enables users to save, track, and manage products across retailers and Google surfaces; and 5) Expanded Android XR with the announcement of audio-first smart glasses, expected to launch this fall." Analyst: Justin Post

@OfficialLoganK @GoogleDeepMind Interesting priorities from Google I/O — faster, cheaper, shinier. But nobody's asking for a more impressive demo. We're asking for a model that doesn't hallucinate when it matters. Reliability > speed. Always.

Welcome to Gemini 3.5 Flash, our most powerful model to date. It pushes the frontier of intelligence, speed, and cost putting 3.5 Flash in a class of its own. We spent the last 6 months making sure Flash is great for real world use cases. It's available everywhere now!

Lately, $GOOGL has been absolutely skyrocketing, pushing my portfolio right back to the brink of breaking even. If only $AXP would finally take off, too!

It looks like there won't be any opportunities for a U.S. stock market pullback anytime soon. I’ve deployed 60% of my cash reserves into $SGOV; my cash position now actually exceeds Buffett's.

It feels as though news regarding the war is no longer capable of triggering significant fluctuations in the stock market; lately, it seems there haven't been any good opportunities to add to my positions.

I'm getting ready to kick off a new research project using Claude's "Projects" feature—just to give it a try.

Strolling through the zoo outdoors in 29-degree weather feels like an act of spiritual discipline in itself; it doesn't feel like spring at all—it is a quintessential summer experience.

Four masters express greater confidence in the growth prospects of these four companies—$GOOGL, $AXP, $AAPL, and $NTDOY—over the next decade; these firms possess particularly deep economic moats. Perhaps this is precisely the foresight characteristic of long-term value investors.

Yesterday, I had an AI simulate Warren Buffett, Charlie Munger, Peter Lynch, and Nassim Taleb to critique my investment portfolio plan. To my surprise, the biggest point of contention among them turned out to be $NVDA

specifically, the unexpected consensus that Nvidia's "golden age" might not even last a full decade.

The biggest worry when traveling during the holidays is that the food at tourist attractions is truly terrible.

It’s a pity that $SPY has already bounced back; my second add-on point wasn't triggered. I share Buffett's view: this decline wasn't deep enough, and the internal risks within the U.S. stock market have not yet been effectively purged.

Anyone who plays *Red Alert 3* knows that preventing an AI opponent from building a superweapon is far more difficult than simply wiping them out entirely. Trump, evidently, has never played the game.

I’ve fully allocated all my remaining capital and finalized my plan to add to my positions based on the closing price of $SPY. Bring on the storm—let it rage even harder!

Hahahahahah