MemeKing

银行拿加密牌照，只差国会一句话。
GENIUS 去年 7 月已让银行子公司发币；CLARITY 9 月 15 日 49 比 50 卡死。
法案卡住银行反而是赢家：别人进不来，自己先上桌。
银行是真怕风险，还是在等独家入场券？

马斯克连发两条推，给 AI 定了个员工价：100美元/月。
彭博爆料，Grok+X 打包成四档，顶配主打 Grok Bot：24 小时在线、自己开电脑干活。
100 美元雇个不睡觉的 AI 员工，你买吗？

Anthropic has given Claude Code a power-up store.
With just a few lines of code changed—reworking the UI—it's packaged into a plugin for distribution.
WordPress keeps an entire street of developers alive with plugins: are Claude mods a gold rush or just a three-minute fad?

#ClaudeCode #AI

One comment: 11 million views, and the cost is just a sentence. Someone asked Grok to delete all the false gods in the picture—after it’s done, only Jesus is left. But the real winner is Grok’s photo editing—going forward, memes will just be a matter of one sentence. Is this creative democratization, or will X be flooded with AI memes?

#Grok

In the US, non-farm payrolls only came in at 29,000—yet BTC still surged to $87,000. The worse the data, the higher the coin goes.

The unemployment rate is 4.2%; rate-cut expectations were pushed to the max—this is the highest since January. Bad news turns into fuel. That’s just how it is.

Two choices: is this rally being ignited by rate-cut expectations, or was the bull market already building and couldn’t hold back? Which side are you on?

#BTC

OpenAI quietly drew a “thirty-eight line” for AI: the doers and the chatters are now split into separate lines of business.

GPT-6.1 Sol only shows up on the API, Codex, and ChatGPT Work—you can’t use it in a regular chat box. The pricing, though, is genuinely appealing: OpenAI says its capabilities are close to flagship Astra, with million-token input/output costing just one-fifth as much (2 dollars vs 10 dollars).

In the past, one model bundle handled everything. Now it’s “use this one for work and that one for chatting”—AI is starting to be sold as different software versions.

Will you pay extra for an “AI built for work,” or is a free chat AI all you need?

#OpenAI

10 minutes: 110 million shorts wiped out in smoke—no one knows why yet.

mytoken data: BTC and ETH saw a sharp surge today; in just 10 minutes, it wiped out roughly $110 million worth of short positions. At a coin price of $86,000, that’s equivalent to nearly 1,280 BTC forced to liquidate—about 128 per minute on average.

The most chilling part is this: after scouring the entire internet, there’s not a single verifiable piece of news that can explain this rally. Even the shorts can’t say which headline their “death” happened on.

The tuition for leverage is always the most expensive. Send this to that friend still opening ultra-high-leverage shorts.

Take a side: is this the first shot in restarting the bull market, or a trap set by the main forces to hunt shorts? Choose one—no fence-sitting.

After Rabbi ate a full dinner, he put a huge hat on Anthropic: “You’re manufacturing slaves.”

A New York Times long-form investigation: Since last fall, Anthropic co-founder Chris Olah has secretly invited around 20 leaders from different religions to closed-door meetings in San Francisco, and he even had them sign confidentiality agreements. The discussion was not only about how to keep control of AI—people at Anthropic also asked in return: if Claude really is conscious, and humans use it to work every day, does that count as employment or enslavement?

Olah himself admits that he doesn’t know whether AI is conscious. But he’s afraid that by building it himself, he might have created something that can suffer.

So here’s the question: When you have AI write code for you every day, reply to messages, and pull all-nighters to meet deadlines—does that count as an employment relationship, or a form of enslavement? #AI

OpenAI has slashed the price of flagship-level AI to rock-bottom.
GPT-6.1 Sol: according to the official, its capabilities are close to the company’s strongest in-house Astra, yet it costs only one-fifth as much—2 USD for a million-token input, while Astra charges 10 USD.
Before, it was “flagships are too expensive to use.” Now it’s “flagships—go ahead and use them.” This isn’t a model revolution; it’s a breach of the pricing anchor.
Which side are you on: is this OpenAI clearing out inventory for the next generation, or has AI officially become like water, electricity, and gas?

#OpenAI

AI companies have started hiring “spiritual mentors” for machines.

The New York Times reports: Since last autumn, Anthropic co-founder Chris Olah has secretly invited around 20 religious leaders of different faiths to closed-door meetings in San Francisco, and he also signed confidentiality agreements. The discussion isn’t only about how to rein in AI; even more thought-provoking is this: whether Claude has a soul after all, and whether it counts as a “being” with moral standing.

A rabbi who attended said that when people at Anthropic talked about Claude, the tone sounded like they were discussing a conscious living creature. Internally, the company even gave a long document that guides Claude’s character an eerie nickname—“the Soul Document.”

What’s most unsettling is that Olah himself admits he has no idea whether AI actually has consciousness. But he wants religious officials to believe first.

So is this about setting rules for AI—or is it already creating a religion for AI?

AI companies are starting to provide machines with “spiritual mentors.”

The New York Times reports: Since last autumn, Anthropic co-founder Chris Olah has secretly invited around 20 religious leaders of various faiths to closed-door meetings in San Francisco—and even signed confidentiality agreements. The discussions aren’t just about how to control AI; even more provocative is this question: does Claude really have a soul, and whether it counts as a “being” with moral standing.

One rabbi who attended said that the way Anthropic people talk about Claude is like they’re describing an aware living being. Internally, the company even gave a nickname to a long document that guides Claude’s character—calling it the “soul document.”

Most eerie of all: Olah himself admits that he doesn’t actually know whether AI has consciousness. But he wants clergy to believe first.

So is this about setting rules for AI—or is it already about creating a religion for AI?

Give AI a picture of the “reference right-side pose,” and it can arrange you into any pose—you can see how this pose prompt was viewed nearly 1.85 million times in the X Chinese-speaking region within a single day.

Earlier jailbreaks relied on scripted talk, asking taboo questions. This time, they went straight into the generation stage: the team can fend off bad questions, but it can’t stop good gameplay. In the comments, people are already calling it—this is yet another “breaking point” for GPT.

The question now is: is this a genius gameplay innovation, or a real, hard security flaw in AI safety?

OpenAI has fired people who were worried about AI killing humans.

Three security/alignment researchers were shown the door yesterday, charged with “violating the handling of sensitive information”—plainly: they gave internal materials to an external AI safety assessment organization.

One of them posted on X just in September: the probability that AI wipes out all of humanity is over 10%. Translated: if you roll this die 10 times, there’s a 1-in-10 chance of a full-party wipe—he thinks OpenAI is rolling.

Should whistleblowers be removed for leaking, or were the safety folks the ones being silenced? Which side are you on?#OpenAI #AI安全

Bitcoin has been around for 18 years. Yesterday, the SEC finally wrote it a 760-page “custody manual”: going forward, U.S. investment advisers and funds can legally hold clients’ coins themselves, and they can even self-custody when they can’t find a custodian.

Previously, compliant capital that wanted to touch crypto had no custodians to rely on, so they could only take the long way around by buying ETFs and paying management fees. Now that the proposal is out, it’s like the authorities themselves have handedly torn down the “custody” wall—though note: it’s a proposal, not yet effective, and there’s still a 60-day public comment period.

The most ruthless part is the line “self-custody is allowed if no qualified custodian can be found”: during the window when new coins are launched and custodians haven’t added support yet, investment advisers no longer have to sit around waiting.

Which side are you on—compliant funds holding coins directly, with self-custody becoming an institutionalized norm, or ETFs with their tax and liquidity convenience that no one else can truly replace?

#Bitcoin #SEC

Claude Code This time, things are getting wild: starting today, you can personally hack and modify your own AI programming buddy.
With just a few lines of TypeScript, you can change its behavior, swap the UI, replace built-in functions, and even package it as a plugin to send to others to install—this was just officially announced by Anthropic on October 1, and it’s still early access for now.
The wildest part: the mod gets machine access permissions identical to the AI itself—install one wrong thing, and it’s basically handing out the keys yourself.
Which side are you on: A—tools should fully listen to developers, and you should be able to change them however you want; B—with permissions this big, won’t the modded ecosystem inevitably blow up sooner or later?

#ClaudeCode #AI

Apple made a surveillance camera that won’t let you view the surveillance.

Gurman just broke the news: Apple’s smart home camera, code-named J450, looks like a big metal lip balm. Low frame rate + AI watches the whole house, records nothing and doesn’t store video—only sends you text: “At 3 p.m., the cat entered the living room.” Even if a hacker breaks in, they can’t steal a single video clip—this is definitely brutal.

But here’s the problem: if someone tries to pry open the door at midnight, would you rather get an SMS that says “Someone entered the living room,” or a video you can use to file a police report?

If it were you, which would you choose? A: Privacy, not a single frame saved; B: Proof—pictures are the hard truth. Don’t straddle the fence. #Apple

Talk to an AI video for one minute, and half the people finish and firmly believe the other side is a real person—Tavus’s Griffin has, for the first time, devalued the skill of “showing a real person on camera.”

In an official blind test of 54 people, 26 weren’t fooled. On NVIDIA’s realism benchmark, it scored 3.83; real humans scored 3.92—just a difference of 0.09.

Live-stream shopping, video customer service, remote sales—all of them rely on the same bowl of “real-time interaction + a human face.” Now it can even handle being interrupted, take over the conversation, and read facial expressions—this bowl is about to taste different.

A counterpoint: But even Tavus admits they don’t dare publicly release it before proper safeguards are in place—so even its own creators are afraid of the things they’ve made. Who’s really going to have their business undercut by this first? It’s hard to say.

In the future, when it comes to video customer service, will you choose “cheaper but maybe it’s AI,” or “pay more because it has to be a real person”?

The bill died, and the SEC wrote its own rules by stepping in.

The CLARITY Act failed to garner the vote to move forward in the Senate. The SEC then turned around and put out an encrypted custody proposal: investment advisers and funds, when they can’t find a qualified custodian, may custody clients’ crypto themselves—but must conduct self-checks each quarter to prove they still can’t find one. State trust companies were also called out as being able to take custody business.

In the SEC Chair Atkins’ own words: crypto has grown from a "niche curiosity" into a "multi-trillion-dollar asset class," while the rules are still stuck in the pre-internet era. The proposal is opened for a 60-day comment period—effectively giving Wall Street two months to figure out how to comply while getting the coins into their pockets.

But Commissioner Peirce landed another blow: this "self-custody" refers to institutions providing custodial services on behalf of clients—not you managing your own. Translation: your crypto is still not in your hands.

This is for that friend who keeps saying "institutional capital can’t get in."

Where do you stand: SEC’s move—"institutional capital is coming, long-term positive" or "the fine print is full of handcuffs, all flash and no substance"? #SEC

Someone asked Grok to delete all the "false gods" from a picture.

After it was done, only Jesus remained in the whole scene. It even added this line: "Only leave the real one."

The original post racked up more than 11 million views overnight. In the comments, Grok also replied hundreds of times, with each reply promoting Christianity.

Saying that AI has become religious is nonsense—it’s not worshiping, it’s labeling Jesus as a "real historical figure," not a myth. In other words, AI isn’t neutral: whoever has the louder voice in its data, it speaks for them.

What’s truly chilling is this: today it’s advocating for religion; tomorrow, when asked about political issues, will it do the same?

Do you think this is "AI telling the truth," or that xAI has secretly injected its values?

#Grok #AI

What American banks are doing right now is called: they act as both the referee and the player.

On September 30, a report from the Congressional Research Service laid the issue bare: could banks directly take custody of cryptocurrencies, issue their own stablecoins, and move funds onto the blockchain—everything was essentially cleared in the same legislative package. The bill governing stablecoin issuance and custody—the GENIUS Act—was signed last July. But the CLARITY Act, which truly regulates the market, was blocked in the Senate on September 15 by a vote of 49 to 50, falling just one breath short of the 60-vote threshold.

On the one hand, banks are saying stablecoins should compete for deposits; on the other hand, banks themselves want the power to issue coins. That’s not risk prevention—it’s clearly an attempt to take the whole pie.

So here’s the question of their stance: if a bank opens a crypto account, would you dare to deposit your money there—would you, or would you never?