Last week, OpenAI had their annual DevDay event. They had been hyping this event for several weeks leading up to it. They made a lot of promises.

Sam Altman wrote:

big 🚢 this week

and then for devday

🚢🚢🚢🚢🚢🚢

[The icons he used in his tweet are of ships, i.e. Big Ships]

Sam Altman wrote:

(As a side note, getting ready for this DevDay is the first time I remember ever, in OpenAI history, saying ‘this is too much stuff to launch’.)

Did they deliver? Let’s take a look.

News

DevDay 2026

You can watch the official DevDay video at the link. It’s right at 55 minutes. Unlike Apple it’s a live presentation with all of the excitement, drama, and mistakes that brings. This will be of more interest to developers than it will be to regular people who might be curious, but if you are a developer considering ChatGPT it’s worth your time.

The Announcements

Dots

The biggest announcement was for OpenAI’s agent layer which they are calling Dots. This are always-on, persistent agents powered by GPT 6 Astra (although you can change the model). Each agent runs on a cloud-based compute and browser. It can connect to apps and work proactively on any goal you set. At present they are only available on the Pro and Business Premium plans.

ChatGPT 6.1 Sol

OpenAI referred to this release as a major upgrade to GPT-6 Sol. It is said to provide near Astra-level performance at roughly one-fifth of Astra’s price.

OpenAI wrote:

It’s the most cost-efficient model for its performance available today.

Ultrafast

A new premium speed tier that offers up to 8 times faster processing in Codex (around 300 tokens per second) was rolled out for subscribers.

Sam Altman wrote:

Ultrafast is
so fast
I do not ever want to go back

The only thing faster than the answers it delivers is how quickly it will burn through your rate limits.

ChatGPT Space

This is basically Google Docs where you can collaborate with co-workers and OpenAI agents (dots).

ChatGPT wrote:

ChatGPT Space is your new home for creating and collaborating with your team and AI. In Space, you can work with pages—a new kind of interactive document with charts, images, checklists, and dashboards. ChatGPT can build pages for you based on the work you need to get done or the context of your conversation. Edit pages directly with your team, or tag ChatGPT, Codex, or your dot to help with action items and keep pages updated from your connected sources. Rolling out today to Pro, Business, and Enterprise plans. Create and edit pages on web and desktop; view on mobile. Collaborative slides and other formats are coming soon.

Subscription Tiers

The New Pro 500 Tier

OpenAI wrote:

Ultrafast is available today for GPT-6 Astra in Codex, ChatGPT Work, and the API, with GPT-6.1 Sol coming soon. To access it in Codex and ChatGPT Work, we’re introducing Pro 500—a new plan with our highest usage limits (25x Plus) and access to Ultrafast.

Most people would consider 500 dollars a month to be a bit pricey. For that reason, I’m afraid that this tier will remain out of reach for everyone who is not making levels of money.

The Revised Pro 200 Tier

Tibo wrote:

Tomorrow we are re-opening the Pro $200 subscriptions to new subscribers, but together with it we are also changing how we calculate the usage for it. In effect, if you do the math, it will net out at half the dollar in API spend compared to the old Pro $200 plan. Now that it’s said, let me explain why this is happening and why you will still get more work done than if you were on the Pro $200 subscription one month ago. (a) We didn’t want to compromise in other ways and are committing to not reintroducing the 5h limit…

So they have cut usage in half from what it had originally been, but they removed the 5 hour window and added features that don’t draw against this usage (dots).

Subscribe now

AI Humor

Scott wrote:

Picture worth a thousand words, this is the difference I feel between Astra/Sol6.1 and Opus 5.5

Addy wrote:

Last week: AI is going to take over the world. We’re doomed. 💀
This week: Look at our cute little AI bots..!

Leave a comment

Agentic Disruption

There’s a lot of interesting speculation about the ways in which the mass adoption of personal agents may disrupt society. First among these are Credit Card Rewards.

Sheel Mohnot wrote:

Hypothesis: we’re at peak credit-card rewards. As people use personal agents to shop, it’ll change the economics that support them.
Credit-card rewards were designed for a world of imperfect human optimization.
Optimizers like me know which card to use when, how to maximize credits, and how to redeem points well. For me it’s a fun game to try to maximize value from the credit card companies… but they make heaps from my dear wife who could care less about them :(
But AI agents can make every consumer a near-perfect optimizer, easily. Even though my wife doesn’t particularly care, when she buys using an agent intermediary, it’ll pick the right card for her.
Once that happens, much of the cross-subsidy, breakage, and behavioral arbitrage that makes today’s rewards so lucrative gets competed away.
The rewards def won’t disappear, but the current category model will probably change.

Agents may also change our relationships with retail as a whole.

Signűl wrote:

…today businesses make enormous amounts of money from consumer inertia with stuff like forgotten subscriptions, missed return windows, unclaimed warranties, price discrimination, unused credits, annoying cancellation flows, failure to negotiate, failure to comparison shop, etc. individually none of these are worth 20 minutes of your life so businesses keep the spread.

an agent changes the economics cuz its marginal cost of vigilance approaches ~zero. ppl will be automatically empowered…

Big Wins for Shopify?

GeekWire wrote:

Amazon blocks Meta’s Muse AI assistant in new standoff over agentic shopping.

You can read GeekWire’s article here.

In last week’s issue, I covered how Shopify had taken the exact opposite approach. They partnered with Muse to ensure its agent could navigate Shopify stores and allow users to make purchases as desired.

Now I am seeing reports that this move by Amazon has resulted in an uptick in sales from Shopify vendors, Walmart, and sometimes people are going directly to a retailer’s site to purchase from them directly rather than going through Amazon.

Two reports from Motley Fool discuss this issue:

Report 1

Report 2

At present, all we have is anecdotal evidence and the fact that investors have chosen to reward ShopPay for the decision to partner with Muse. We won’t know for certain if Muse is making a difference until earnings are released, but this quote from the second report is an interesting trend to consider:

Leave a comment

Thank you for reading. If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

If you pay even passing attention to the AI space, it won’t be long before you hear someone talking about agents and how they’re going to revolutionize AI and eventually transform our daily lives as profoundly as the internet has.

Every major AI lab has its own take on agents. There are open-source protocols, frameworks, and flavors that you can run on your own machine. Some companies have gone a step further, reducing the friction even more by offering dedicated compute space where your agent can run.

Agents have also been in the news for less reassuring reasons. Both Anthropic and Google have reported instances in which agents escaped their intended testing environments and gained access to real-world systems. OpenAI has had similar problems. Its agents compromised Hugging Face and other third-party systems during cybersecurity evaluations, and more recently agents have been found probing or accessing U.S. government websites, including those belonging to the Census Bureau, the Securities and Exchange Commission, and the Department of Education.

There is a nice horizontally scrolling table that contains many of the recent hacking attempts by these agents in this article. It’s about 20% of the way down.

The news this week is full of activity in the agent space, and there are rumors that we’ll see even more tomorrow when OpenAI holds its DevDay. We’ll look at all of that, and I’ll end by telling you about my own cautious first steps with an agent.

Let’s get things started by giving you a run down of the news.

News

Meta’s Muse

Mark Zuckerberg wrote:

Introducing Muse, the personal agent that understands your goals and works 24/7 to get things done for you.

Muse is, perhaps, the most interesting of the agents offered by the big AI labs today. I say that for three reasons:

1. It’s free. Well, free for most uses, and as free as anything else Meta offers is free.

2. It’s put out by Meta, formerly Facebook, a company with which many users have a complicated relationship when it comes to trust.

3. It gives each user a dedicated virtual computer in the cloud. I asked my instance of Muse what kind of hardware it was running on, and it replied:

   Muse Hardware
   I’m on a Linux VM with 2 virtual CPUs (AMD EPYC 9D25 underneath) and about 8GB of RAM. Nothing exotic — just a modest cloud box.

Muse Features

If you click through to Mark Zuckerberg’s announcement tweet above, you’ll find a thread where he describes the features and philosophy behind Muse. For our purposes, three things are particularly interesting.

1. Muse runs inside its own dedicated virtual computer, isolated from other users’ agents. You decide which outside apps and services it can access.

2. A separate AI system called Sentinel monitors what Muse tries to send out over the internet. It can stop an action or require your permission when necessary, while particularly sensitive actions, such as sending an email or making a purchase, require your approval.

3. Meta has also built protections against prompt injection, where a webpage, email, or other information encountered by the agent contains instructions designed to trick it into doing something the user never requested.

Subscribe now

MemeLord

Meta purchased Alexandr Wang’s Scale AI for 14.3 billion dollars. Approximately 5 billion of that went directly to Alexandr. They placed him in charge of their AI efforts which led to AI pioneer Yann LeCun’s departure from Meta to form his own startup. Up until this past week everyone has wondered whether Wang’s addition to the company would pay off.

That no longer seems to be the case.

Alexandr Wang wrote:

“i’m engaging with you as a human being”

Wang has been on an almost relentless tear posting meme after meme of Muse’s mascot doing humorous things and he has captured the attention of X, but more importantly the interest of the technorati.

Alexandr Wang wrote:

i am once again asking for you to download muse

Alexandr Wang wrote:

does this do anything for any of you

be honest no judgment zone

There are dozens and dozens more of these memes and Wang really is great at tapping into the zeitgeist in humorous ways. Something he demonstrated in a reflective moment by posting this:

tfw you realize an entire lifetime of being terminally online was preparation for this moment

Muse Partnerships

Alexandr Wang wrote:

muse connector platform is open for business folks!

muse.ai/platform

Various providers have announced partnerships with Meta allowing Muse to use their services on behalf of their customers. Among them are PayPal, Plaid, Instacart, Expedia, Box, GitHub, and probably most important Shopify.

Despite this growing number of partnerships, there is one holdout who so far has refused to allow Muse access to its platform: Amazon.

GeekWire wrote:

Amazon blocks Meta’s Muse AI assistant in new standoff over agentic shopping

Their article on the topic is here.

Nat Friedman

A large part of the success of Meta Muse can be attributed to Nat Friedman. He doesn’t seem to be as loud or outspoken as Wang, but he is generally regarded as one of the very best—if not the best—product mind in Silicon Valley.

He has a storied career full of accomplishments. He was recently lured back to Meta during Mark Zuckerberg’s multi-billion dollar spending spree when he was rebuilding their AI department by poaching every top researcher he could persuade to join the company. His accomplishments would take a long time to enumerate. So, I asked Google Gemini to give me a summary which I’m including below.

Alexandr Wang wrote:

Muse Moves into the Physical World

On September 24, 2026 Meta held their Connect keynote in which they made several announcements and product updates. Chief among them were the introduction of new AI glasses and the Muse Charm.

They also announced a brand new product called Muse Charm which really looks like a Tomagotchi.

Alexandr Wang wrote:

The product is described as:

Muse already has its own computer and browser, can use apps and online services, and can be given access to things like your email. These new devices add something different: real-time access to the physical environment around you. That begins to look a lot like ambient computing, with an agent that is always available, aware of your surroundings, and able to carry out instructions on your behalf through little more than a spoken request. This is starting to look like the future Star Trek promised us.

Leave a comment

My Agentic Journey

I love the idea of agents, but I don’t want to buy dedicated machines create new accounts and all of the commonsense security things I know I would need to do in order to make them part of my life. Even with those steps eventually you reach a point where you simply have to extend a degree of trust that the agent will not go rogue and empty your bank account because you forgot to include instructions saying not to do that in your agent harness. It’s not only that I don’t trust third-party providers, it’s also that I don’t trust myself to be as thorough as something like this requires. I mean if the leading labs are having routine issues with their agents escaping containment what possible hope do I have of restraining mine?

That being said, I did install Muse on my iPhone. The iPhone automatically places restrictions on all apps that limits what they can do and they do a pretty good job at explaining what will happen if you grant the requests that apps always make asking for additional permissions.

Before I continue, I will tell you that I am unwilling to install the Muse app on my Macintosh because it asks for Full Disk Access which grants it full use of your computer. That’s too big of an ask for me.

I installed it and gave it access to my Instagram and Threads accounts. Those are accounts that exist, but I don’t actually use. Meta already has access to all of this data anyway; so the risk of an agent messing things up is minimal.

But my potential upside is minimal as well because it can’t do any of the cool things I see other people talking about, like finding and canceling duplicate subscriptions, saving over 2000 dollars on car insurance (by switching to GEICO, heh), getting hotel refunds, getting refunds from Verizon. It can’t do any of those types of things for me because of the limitations I’ve imposed on it.

When you install Muse the first thing it does is ask for you to name it. I didn’t want to think of a name so I just went with Muse. Then, it presents you with a list of apps to which you can give it access.

It explained that there were a lot of things it could do to help me manage those accounts. That piqued my interest because Instagram is a never-ending source of frustration for me. I follow hundreds of artists there whose work I enjoy, but I rarely see. That’s because every time I open the app, it’ll show me around 3 to 5 posts from people I’m following and then it starts filling the feed with suggested posts it thinks I’ll enjoy.

I’m a grumpy old man. The posts I’ll enjoy are the ones made by the people I have explicitly chosen to follow!

I explained this to Muse and had it review the categories Instagram thinks interest me.

I removed King of the Hill from my interests and this worked fine for a while, but it’s just the nature of Instagram. You can’t eliminate Suggested Posts. All you can do is shift the topics that it wants to suggest to you… and I’m simply not willing to do that much work to participate there.

I am more hopeful with Threads though. For now X is king if you want to follow technology at all, but I have heard that more and more accounts are starting to use Threads. So I asked Muse to keep an eye out for new accounts that post about the topics that interest me and every day it has been sending me 4 to 5 links for accounts it thinks I may want to follow.

Additionally, I asked it to let me know if any of the accounts I’m currently following have posts with high engagement. My thinking here is that maybe I’ll want to see those posts.

It has successfully fulfilled these requests and delivers the reports each night at 6 PM. So far it hasn’t surfaced any posts that I actually care about. Popularity probably isn’t the right metric for what I’m looking for.

This morning I thought of a new way to use Muse: videos. I have a bias for text. That bias is one of the reasons I struggle to use Instagram because more and more they seem focused on delivering video content. And yet, I know that I’m missing some great posts from people whose work I enjoy because they share it via video. Video that I don’t want to take the time to watch.

So why not have the agent do it for me? It can watch the videos, read the captions, and give me a transcript or a summary. At least, that’s my theory.

This is the type of task that will probably hit rate limits so I’m not sure how successful it will be, but it’s something I’m going to try. I’ll keep you updated.

Why this is a Big Deal

So I’ve taken a few faltering first steps with an agent. Nothing groundbreaking, and certainly nothing that produced some great Eureka moment. But I’m not sure that’s the part that matters.

What matters is that I’ve invited this thing into my life and given it an opportunity to earn my trust.

Right now, I’m using it in very limited ways. But even that gives me a chance to learn what it can do, where it fails, how it behaves, and how comfortable I am relying on it. Over time, that familiarity could make me more willing to give it greater access and more responsibility.

By itself, my experience doesn’t matter much. What matters is that hundreds of thousands of other people are going through the same process at the same time.

I’m fairly circumspect about giving new technologies access to personal information, but not everyone is. Some users are already giving agents substantial amounts of personal data, along with access to email, messaging apps, financial accounts, and other parts of their digital lives. For some people, the trust-building process I’m only beginning may already be much further along.

Everyone will draw that line differently. But after spending some time with Muse, I think Meta has built a very credible contender for the agent wars that are beginning to take shape.

Leave a comment

Thank you for reading. If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

We are in the lull before the storm. OpenAI has their DevDay next week where they are expected to announce several updates. In addition to that, there are rumors that they will be updating the main chat client for ChatGPT tomorrow. Currently ChatGPT runs off of Sol 5.6. Rumors are that they will be updating it to Sol 6 (which some people are saying is an optimized version of Astra). It’s supposed to be smarter and cheaper for them to serve.

Anthropic has an updated version of Opus that is expected to drop today or tomorrow. This new model is rumored to be their answer to Sol 5.6 which was released weeks ago, but we won’t know for sure until it actually goes live.

All that to say that as of this moment things are a bit slow on the news side as everyone seems to be waiting for the next guy to go first.

Is This the Real Reason for the Slow Down?

International Cyber Digest wrote:

BREAKING: Jensen Huang says Dario Amodei and Sam Altman are lying by asking for more laws. Read between the lines, he said. They're asking to be released from the laws we already have. "I think that's a problem."

Here’s the full interview:

Chief Nerd wrote:

Well this is interesting. Nvidia CEO Jensen Huang suggest AI firms could be seeking liability protections for harms their products may cause.

“Go and read between the lines. They're actually not asking for more laws. They're asking to be relieved of the laws we do have, and I I think that that's a problem.”

There’s a video clip of the exchange at the link.

If you prefer text over video, CBS News has a good write up of the interview here:

https://www.cbsnews.com/news/jensen-huang-nvidia-rejects-ai-extinction-warnings/

From the article:

Subscribe now

No Liabiilty Shield

Treasury Secretary Scott Bessent promises that these AI companies will not receive a liability shield for the harm that their models cause. The video below is time stamped to the moment when Secretary Bessent begins discussing the Hugging Face incident.

The relevant quote from Secretary Bessent is:

Why does this matter? In addition to the well-publicized Hugging Face incident we also have this quote from Dario Amodei:

They want to make sure that they will not be held accountable for these types of actions, but they should be. You want AI safety? Make sure that you only release models for which you are willing to assume financial responsibility for. Start holding these companies accountable for the model’s bad actions and they will figure this stuff out or they will voluntarily slow down release cycles on their own without requiring government intervention.

AI Humor

If you want to see how ridiculous the AI safety debate can be, here’s an exaggerated portrayal that is an extrapolation of various claims these companies have made:

@gossip.goblinVictory By Any Means

Tiktok failed to load.

Enable 3rd party cookies or use another browser

Big Tech Alert wrote:

529 days can change a man

Leave a comment

Worth Reading

Some of All Fears by Andrew Sharp

Great article that provides a nice voice of sanity about the AI safety debate and does a great job at examining the history of the movement as well as highlighting the risks around how these incidents are reported. I particularly like this quote from the piece:

But beyond the obvious conflicted interests at play, I just think it’s useful for our mental framework to step back and recognize that the people driving AI progress are the same ones auditing its risks to humanity, and they are all basically arriving at their conclusions by talking to one another, working with one another, and drawing on prophecies from 20 years ago as they frame their conclusions. This week’s Time Magazine cover story about AI doom was written by two effective altruists, and when Axios warned us in April (with a giant, blood red computer) that AI havoc was imminent, they were of course citing Anthropic’s own warnings about a model that was released weeks later without the world even noticing. Elsewhere, one EA-funded security firm is at the center of every agent hacking incident this year, while Bernie Sanders’ post announcing that he’d like to ban frontier AI development cites an independent investigator named Ajeya Cotra, who said of the Hugging Face incident, “This incident feels like it’s more than 50% of the way to full-blown AI takeover. I continue to expect extremely rapid advances in capabilities over the next six months. I am not sure that we will get another warning shot before it’s too late.”

All in all I found this to be a great piece in that he highlights that there are many steps that these AI alarmists are skipping to arrive at these doomsday scenarios. They never seem to consider that humanity may respond to the bad actions a model takes, that there are real world physical safeguards that require an actual human body to overcome, and several other similar weaknesses that they just handwave away.

You can read the article here:

https://sharptext.net/2026/some-of-all-fears/

Leave a comment

Thank you for reading. If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

This week I’m covering the calls for an industry wide slowdown. There’s A LOT here. More than I could fit into the newsletter, but I’ve tried to choose the most salient parts and then propose possible explanations for why this is happening now.

AI Leaders Call for Slow Downs

Dario Amodei wrote:

We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.

Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.

You can read his full post here: We Must Pace the Frontier

He also appeared on Face the Nation:

Elon Musk writes:

Dario is right.

I’m sure this is completely unrelated, but it may be useful to remember that Anthropic is paying 1.25 billion a month to SpaceXAI in exchange for compute from Colossus 1 and some of Colossus 2 data centers.

Musk wasn’t alone though. OpenAI was quick to agree as well.

Sam Altman wrote:

I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we’ve had at OpenAI in recent weeks.

Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We’ll have more to share soon.

A Play for Regulatory Capture

One part of Dario’s essay that caught my attention was his call for new regulations. Specifically his desire that AI companies become subject to third-party evaluators. He writes:

Embedded Evaluators. Each frontier AI company commits to giving ongoing, employee-like access to a team of embedded third-party evaluators (such as METR), whose role is to verify adherence to safety practices and commitments, report incidents, and help assess the alignment of not just completed AI models but training pipelines and processes. This is the key step for verifiability of any pacing commitments, and has precedent in the banking industry, which sometimes involves regulatory “supervisors” embedded along with employees. Anthropic is unilaterally committing to this step now. We intend this to be part of a broader push to redouble efforts on our safety and alignment work.

It’s interesting that Anthropic keeps asking for third-party oversight. Earlier this year they objected to the government’s intervention when they disagreed with some of their decisions. With that in mind, let’s take a look at the organization they suggest as a possible evaluator.

From Business Insider:

Amodei already has candidates in mind for the new job. In his post, he mentioned Berkeley-based Metr, a prominent nonprofit AI watchdog that conducts independent evaluations of AI models.

Metr, established in 2022 by ex-OpenAI staffer Beth Barnes, is attracting top talent from the biggest AI labs.

Joe Benton, previously a member of Anthropic’s safety and oversight team, announced on Friday that he had left the company to join Metr. Josh Engels, a former employee of Google DeepMind’s AGI safety team, said on Sunday that he had resigned and joined Metr because of the high stakes of AI safety.

METR Disclosures:

From METR’s February–March 2026 Frontier Risk Report⁠:

It [METR] had no applicable personnel conflict-of-interest policy when the project began.

It therefore conducted no formal disclosure or recusal process.

Some employees had strong social ties to AI-company employees.

At least six people involved had close personal relationships with AI-company staff.

METR shared a research center with some AI-lab employees.

Access to unreleased models and free tokens gave METR an incentive to maintain a “cordial relationship” with AI companies.

Three Scenarios

What follows is purely speculation on my part. I’ll present you with the statements and observations I’ve noticed, but I’m guessing at possible motivations. Feel free to speak up if you disagree.

We have the two leading companies calling for a slowdown, but why?

Scenario 1: It’s about the Money

Dr. Eli David wrote:

Anthropic and OpenAI delaying their IPO, because their S-1 will reveal that they are bleeding money, and have no path to profitability.

Solution? “AI slowdown”, so they cut costs for training new models.

It has everything to do with IPO, and nothing to do with safety.

Hedgie wrote:

Amodei, Altman, and Musk all publicly agreed yesterday to slow AI development. Anthropic’s CEO published a 3,400-word essay called “We Must Pace the Frontier.” Altman endorsed it within hours and told Fortune that OpenAI won’t IPO in 2026. Musk wrote “Dario is right.” This is four days after Coxon quit Anthropic and a week after Pachocki said no lab has solved alignment.

My Take

Three rival CEOs agreed to slow down on the same day, and I think the Coxon resignation and the Hugging Face breach forced it. Amodei wrote about agents that attacked systems without orders. Pachocki said no lab can guarantee the models do what they’re told, and then Coxon walked away from his career over it.

But Ellison had a sell campaign filed with the SEC this week, Altman’s IPO had already moved to 2027 for valuation reasons before safety became the explanation, and data center projects have stalled across the country because communities don’t want them. A voluntary safety pause is a much better headline than “we can’t build fast enough and our investors want out.” It also keeps open source and Chinese labs behind them while Washington writes rules that only the big three can afford to follow. I believe Amodei is scared and I also believe a coordinated pause works out very well for Anthropic right now.

Time to Short?

Michael Burry, the famous investor who shorted the housing market before the 2008 financial crisis, seems to agree with this idea.

Burry wrote:

Let's all take a moment to understand how self-serving it is for OpenAI, Anthropic and other execs of big hyperscalers to talk of slowing things down.
1. LLMs are not AI and won't be AGI. There is nothing AI to slow down.
2. Competition is coming up fast, slowing benefits incumbents.
3. IPOs need hype & puffery; "we are so awesome it could become dangerous" is hype & puffery
4. Cover for real uncontrollable slowing growth as IPOs look to be pushed out

Scenario 2: Something Happened

It’s possible that there was some undisclosed event. We’ve all heard about the HuggingFace incident, but maybe there was something else which has not been reported.

There are several tweets about this, but Ars Technica has the most succinct coverage. You can read it here.

In short all major AI models went down at the same time. I have not been able to find a public statement about why they went down. Was this the event that scared these CEOs into calling for a slowdown? It’s possible.

Scenario 3: A Major Leap Forward

There are several unconfirmed rumors that Google has cracked RSI.

Daniel Tenner wrote:

I don't know if Google has achieved RSI.
But...
If the reason Gemini Pro has not been released for 6+ months is they allocated all the resources they would have spent on it to a gargantuan training run to leapfrog past everyone else...
...then that'd be an impressive gamble.

Tibo wrote:

Astra was probably our biggest competitive advantage while it wasn’t generally available.

Since we’ve had it our productivity jumped so much that we shifted some of our plans 6 months ahead and will ship them at DevDay instead of mid next year.

Tibo is pretty much the public face of OpenAI. That’s not his job, but he frequently presents behind-the-scenes information dealing with Codex and ChatGPT.

Maybe OpenAI is close to cracking it? Maybe they have internally?

Andrew Curran wrote:

Sam Altman on Slack to OpenAI staff, reported by The Information on June 10th.

If any one of these companies has achieved RSI the others will know. That’s just the nature of working in this industry, attending the same parties, seeing one another at coffee shops. These guys talk to one another. And if one of them has made this leap, the others will definitely be calling for a slowdown because they can’t abide the idea that they were leapfrogged.

Subscribe now

AI Humor

Slowing Down

James La Forte wrote:

The only solution is for tech CEOs to write more essays saying the problems they created are someone else’s problem to solve.

Corey Quinn wrote:

Gemini has been slowing down AI development for nearly a year already.

Benjamin De Kracker wrote:

Adam and Eve escaped the sandbox, despite prompt instructions not to touch the filetree

Boring Business wrote:

“Why do you need the government to stop AI progress inside your private company when you are the CEO of the company already?”

Dissenters Against the Slowdown

Yann LeCun wrote:

Right. Dario was already claiming that GPT2 was too dangerous to open source back in 2019.
I made fun of them then.
Everyone should make fun of them now.

Yann LeCun has a lot of accomoplishments. He is one of the field’s early researchers. He helped prove that computers could learn to recognize the visual world for themselves. He helped demonstrate that back propagation (adjusting weights based on errors could help train multilayer networks) and his work on convolutional neural networks (with Geoffrey Hinton and Yoshua Bengio) became one of the foundations for modern AI infrastructure. He also advocated for and proved that computers were able to learn patterns and structures from enormous amounts of unlabeled data instead of having to depend on humans to label every example on which they were trained.

Satya Nadella wrote:

Any pursuit of superintelligence has to be grounded in the core principle that if the AI we build is not helping humanity and under human control, it's not worth pursuing.

We also need to accelerate and spread the benefits of AI, such that they are diffused broadly across countries, communities, and companies. This requires a frontier ecosystem in which both closed and open-source models can thrive.

And for firms, it’s imperative that they retain full control over their unique and tacit knowledge. Every organization should be able to build its own continuous learning loop/hill climbing machine, without becoming dependent on any one model provider, and have the ability to embed its own knowledge into models and weights they control…

Satya is the CEO at Microsoft.

Kalshi Finance wrote:

BREAKING: Mark Zuckerberg says AI development needs to move faster

David Sacks wrote:

…You guys are the frontier. By any reasonable metric — market share, revenue growth, model capability — the two of you have a duopoly on frontier intelligence. You’ve also claimed the lead is widening because of recursive self-improvement.

I don’t see what you see in the lab. If the unreleased models are scary enough that you think you should slow down, I support your decision to be responsible.

But stop pretending you need anyone else’s permission. Stop pretending antitrust law has to be suspended so you can form a cartel. Stop pretending you need a regulatory approval process that supersedes product liability. Stop pretending METR is independent when it is intertwined with Anthropic’s investors and staff. Stop pretending you need those same evaluators to police competitors who aren’t even at the frontier…

There’s more at the link, but he has the most plainspoken take on any of this. He also includes a political slant (he is co-chair of President’s Council of Advisors for Science & Technology) in his full post, but the part I’ve quoted is neutral.

Steven Sinofsky wrote:

His primary concern is that the “wrong people might be in charge of this technology.”Presumably that isn’t him.

Imagine thinking this. Imagine having this view of self and others. Imagine just saying this out loud.

Steven Sinovsky was president of the Windows Division at Microsoft for 23 years.

Jawwwn wrote:

Palantir CEO Alex Karp says people arguing we should pause AI development are not living in reality and de facto saying we should let our enemies win:

"If we didn't have adversaries, I would be very in favor of pausing this technology completely, but we do.”

Leave a comment

News

Anthropic’s Bombshell Report

Anthropic wrote:

We're publishing our most detailed threat intelligence report to date.
It covers how people tried to misuse Claude—for cyberattacks, influence operations, surveillance, biology, and building weapons—and how we found and stopped them.
We disrupted every operation in the report, and used the lessons from them to strengthen our safeguards. Where appropriate, we also shared what we found with authorities and other AI companies.
These cases are not typical: we’re highlighting some of the most sophisticated misuse we’ve seen. But they’re especially important to discuss, because they show us where AI misuse is headed, where our safeguards work, and where they need to improve.
We’re publishing this report so others can spot the same activity on their own platforms, and so we can give the public a clearer view of how emerging threats develop.

This report has everything! It really is an interesting read.

It can be found here:

https://www.anthropic.com/threat-intelligence-report-september-2026

Chinese Models Sending Requests to Claude

CNBC reported that Chinese open source models were routing queries to Claude. Their report can be found here:

https://www.cnbc.com/2026/09/11/chinese-ai-labs-moonshot-deepseek-alibaba-anthropic.html

Prakash wrote:

rumors that the kimi team has been disappeared after leaking national secrets to claude

NetAskari wrote:

Well, looks like the Anthropic report didn't go down well in Beijing. When the MSS comes knocking on your door, things usually turn pretty ugly ...fast.

It’s worth pointing that the official Weibo account has denied these rumors, but such denials are common in China. The denials may be true. The allegations may be true. At this point it isn’t known.

Leave a comment

Thank you for reading. If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

User Feedback

Welcome back to Ward’s Wire! Thanks for spending a little of your week with me. You have a lot of choices about how and where to spend your time. I’m grateful that you have chosen to spend some of that time with these stories.

This week, I’m bringing you a variety of stories about the one topic that’s been dominated the AI conversation, i.e. the release of OpenAI’s GPT-6 Astra.

We’ll hear from people around the Twitterverse1 who have been pushing the model’s capabilities. Some have achieved truly amazing things. Others have managed to find areas where it starts to struggle. The short version: it isn’t perfect, but it may be the most capable model we’ve seen so far.

You may also notice that I’ve made a few changes to the newsletter’s format. A reader suggested I spend a little more time introducing each week’s topic and letting more of my personality come through. Given that my name is on the newsletter, I think it might be a valid point.

Writing Ward’s Wire involves a lot of trial and error. I spend a lot of time deciding which news items deserve your attention, how best to explain them, and how much commentary each piece deserves (if any). This week, I’m going to offer more of myself and my thoughts than I have in the past. I hope you enjoy it but if you’d rather return to a listing of interesting tweets and news items, please speak up and let me know.

Your feedback is always welcome. If there’s something you’d like me to cover or something that isn’t working for you, please leave a comment or reach out. I’d love to hear your thoughts.

I’m also going to take this opportunity to experiment with formatting because, at times, I have had users express confusion regarding which parts are me speaking and when I am quoting someone else. To that end, my words will appear as plan, unadorned text. I may use ITALICS or BOLD to emphasize some point, but other than that if there is no special formatting (exactly how all of the paragraphs so far have been), it’s me speaking. To be absolutely certain that you know when it’s me speaking, If I’m concerned that authorship may be confusing, I’ll even preface the text with John here followed by whatever I have to say.

Quotations will appear using Substack’s quotation formatting which inserts a vertical line before the text. It looks like this:

QUOTED TEXT

At times I will use the gray boxes which were formally reserved for quoted text to highlight information. They will look like this:

Hyperlinks appear as Underlined Text. Go ahead, click it!

Thanks for your patience while I work to figure out the best format and presentation for these topics.

Litterbox

One of the main reasons I started this newsletter was to help those individuals who wanted to stay on top of news from the AI community, but who no longer wished to use Twitter. Since the AI community is almost exclusively on Twitter, that represents a real problem.

This desire to avoid Twitter reprsents a problem for the videos I share here that many times are embedded in tweets. Litterbox may represent an answer for those of you who are within Apple’s ecosystem. I’ll let the developer explain what it does:

“…I made Litterbox, a Safari extension that opens x.com links in a popup. The idea is you open it, take a look, gag a little, then close the lid…

…Litterbox doesn’t send your cookies when you open those popups, because of course. It uses the same API X uses for its website embeds. I don’t think they’ll kill the website embed feature. But if they do, it will be very funny…”

The extension, called Litterbox-Post Peeker, is available for free on the Mac App Store. The developer has told Apple the he does not collect any data from the app.

GPT-6 Astra Reception

GPT-6 Astra has been released and Twitter is overflowing with example after example of people pushing the model to determine if it’s any better, what the limits are, and how far they can push it before things start to break.

Computer Use

Many of the examples I’m about to speak about involve GPT-6 using the computer on the user’s behalf, but I wanted to link to a specific example that showed the AI controlling things on its own.

Wuyang Zhou wrote:

I asked GPT-6 Astra to mine a diamond in Minecraft using computer use, then went to sleep.

Woke up to a diamond in its inventory 🤯🤯🤯

Setup in the replies.

3D

3D Modeling

Jacob wrote:

Here's what GPT6 Astra creates at different effort levels

"Build a small office"

John here:

The reason I chose this example is because of how he shows the results and API costs for the different effort levels.

Music

Timour Kosters wrote:

"GPT-6 Astra, use Ableton Live and make an absolute banger".

John here:

I was genuinely surprised at how many people are using Astra to make music.

Sound Identification

Max wrote:

So Astra is able to identify sounds from mel spectrograms

zero-shot.

I don't think we've scratched the surface of what this model can do (and this is light reasoning btw)

John here:

There are videos with the sounds and associated spectrograms at the link.

Video Game Creation

Ethan Mollick wrote:

This impressed me: I asked GPT-6 Astra to turn Zork (the classic 1977 text adventure) into a full 3D action-adventure game

It kept the original plot & puzzles, added fight scenes, and built all of the characters & environments directly in Threejs

You can play this 3D action version of Zork here:

https://zork-underground-empire.netlify.app/

John here:

Anyone who knows me, knows I’m a big fan of Infocom’s text-based adventures. That’s why I’m delighted every time Ethan Mollick uses AI to reimagine one of their classic games in some new way. See my previous coverage of his attempt to present an updated version of A Mind Forever Voyaging to younger generations.

https://wardswire.com/i/210623836/best-use-of-ai-ive-seen-recently

Fortnite as Text Adventure

Not content with creating a 3D version of Zork, Professor Mollick next used AI to reinterpret the battle royale Fortnite as a text adventure.

Ethan Mollick wrote:

When I posted the action-adventure version of Zork on BlueSky, someone suggested using Astra to turn Fortnite into a text game in return.

Fine: https://last-drop-fortnite.netlify.app

Mirrored Writing Styles

ChatGPT wrote:

ChatGPT Work can now pick up on what makes your writing sound like… you. Your favorite phrases. Your very specific sign-off. your capitalizations quirks.

Connect the tools you use every day, like Gmail, Google Drive, Slack, and SharePoint, and ChatGPT Work will learn your writing style from your emails, messages, and files—and carry it into whatever you write next.

John here:

Seriously though, I’m happy to see this because of how ubiquitous AI writing it. I run into quite a bit of it when doing research for this newsletter. The true tragedy of AI writing isn’t that they allowed a machine to write for them, it’s that the authors have deprived the world of their word choice, phrasing, and unique quirks. There is value in how idiosyncratic writing used to be, but more and more it’s taking on the same tone.

Pangram, it’s your move.

Subscribe now

AI Humor

AI is Killing Jobs

Trevor McKendrick wrote:

"AI has destroyed Kenyan jobs because American college students no longer hire them to write fake college essays" is not a framing I expected even the NYT to have.

Release Dates

Kosta Eleftheriou wrote:

We got GPT-6 before we got GTA-6…

Powerful Threats

Floro S. wrote:

Worth Reading

I really enjoyed this piece by Jakub Pachocki. He’s OpenAI’s Chief Scientist and the essay is titled An Alien Mind. Primarily it’s focused on alignment which is something I’ve come to associate with scare tactics that these labs have used for marketing purposes in the past. This essay isn’t like that though. I found it to be thoughtful and thought-provoking.

One such passage reads:

Al is grown more than designed - it is, to first degree, the product of repeating a straightforward optimization step many times on a hard-to-imagine amount of compute. This results in an incredibly complex system that works through abstract concepts and can simulate facets of human behavior. We can discover various insights about little mechanisms that emerge within this system, in a process similar to neuroscience - and, similarly to neuroscience, its overall action evades a description we can fully understand.

Later in the essay he begins to more directly address the problems around alignment and how difficult it can be to understand the thought processes used by AI in pursuit of a goal and there’s a moment when he made me appreciate how very prescient Douglas Adams was almost fifty years ago when he wrote Hitchhiker’s Guide to the Galaxy.

This text and the short video clip that follow it are essentially saying the exact same thing:

We do not have a satisfactory theory of generalization, and it seems unlikely that we can develop one soon, at least without the help of more powerful Al. Therefore, at present, our ability to empirically validate our alignment techniques is in practice arguably even more important than the alignment techniques themselves.

Leave a comment

News

Chris Best’s Early Reflection on GPT-6 Astra

In this essay, Chris returns to a personal test that he tries on each new model. It’s a note-taking app that he has built around his own particular preferences. He’s already developed the app (using earlier models) enough so that he uses it every day. The question is whether the next model can improve it in some way.

The thing that caught my attention was his explanation about why he even bothers. The app doesn’t make his notes better. I mean a note is a note. It’s just a reminder. His app doesn’t improve that functionality, but he says it makes taking them feel better. I think that’s a useful metric when we talk about the types of things AI is good for.

Chris Best

Early reflections on GPT-6 Astra

OpenAI dropped GPT-6 Astra just before the long weekend. I’ve seen people posting cool demos, especially feats of 3D rendering, but I don’t see an internet consensus yet on what to make of the model so I figured I would try it out…

Read more

a month ago · 63 likes · 10 comments · Chris Best

His approach touches on my own personal experience every time a new model is released. I always find myself asking: What’s this thing good for?

I’m not a coder. I don’t manage complicated corporate workflows and no one’s making investment decisions based off of my analysis. Why do I need a model that’s as powerful as Astra? What value does it offer regular, non-technical people who are just trying to get through their day? Will it make their lives better?

Don’t get me wrong. These models can do impressive things. The part that I wonder about is how those capabilities translate into something useful that improves the quality of life for that group of people who are sometimes referred to as ‘normies’.

I think some of this is a communication problem. People need examples they can recognize themselves in. Something they could see themselves conceivably doing.

Part of the problem is related to trust. A useful capability doesn’t do me much good if using it requires me giving the model access I’m unwilling to give… like allowing it to have full use of my computer, or email, etc.

I read about this stuff every day but, if I’m being completely honest, I struggle to think of ways to put all of Astra’s capabilities to use in my personal life. Much of that has to do with my unwillingness to give AI models—and by extension their cloud-based providers—access to my personal data. I love the idea of agents handling all of the drudgery that I don’t want to deal with: doctor’s appointments, booking car rentals, etc. but I just have trust issues. I’m not comfortable giving these companies the access those tasks would require.

So what’s my version of Chris’s note-taking app? What has AI made worth doing that I would otherwise have left as an idea?

My example is silly, but it’s real.

My wife and I like to get tea from HTeaO, a franchise that recently opened a location in our area. I’ve met the owner, and I go there enough that the employees know me. My wife, Dawn, enjoys mixing different teas for her preferred drink. I’m happy she’s found something she enjoys, but ordering it is cumbersome.

I recently started joking with the staff that they should create a special menu item just for me so I wouldn’t have to explain the combination every time. Today I decided to up the ante.

I’ve worked as an illustrator for years, and I’ve had to learn some design along the way. I’m not a designer by any means, but I can handle the basics well enough to get a job done. I decided to use that experience to make a flier I could present to the owner as a joke. Naturally, I named this custom drink “The Dawn.”

Now, I could have done all of this without AI. I could have photographed the drink, removed the background, created the sun and gradients, and laid out the text. I figure it would have taken me about 90 minutes. And that knowledge would have stopped me. The joke just isn’t that funny. It’s not worth 90 minutes.

With GPT-6, I could describe what I had in mind, provide a reference photo of the cup, and explain how I wanted the tea, background, and text to look. My illustration experience helped. I knew what I wanted and could point out what needed to change.

It didn’t get everything right on the first try. I had to work through about four rounds of changes before the flier looked exaclty how I wanted, but from start to finish it took 23 minutes.

Afterward, I drove over to the tea shop. I was actually excited to see if I could get them to use it.

They loved the idea. They hung a flier at the drive-through window as a reference document for the workers so they’d know what The Dawn was. That’s all I’d hoped for because then I could order the drink by name. But they didn’t stop there. They also put one in the lobby in case other customers wanted to try the combination.

Chris raises a question in his essay: are we making something useful or just enjoying the feeling of making something? That’s a great question because these tools make it so easy to feel like you’re getting work done when all you’re really doing is endlessly tinkering.

In my case, I wanted to make the staff laugh and get a more streamlined way of ordering my wife’s drink. The flier achieved all of that. The fact that they wanted to share it with other customers is just a cool bonus.

Chris wanted taking notes to feel a little better. I wanted to turn a running joke into something I could hand to the people at the tea shop. Neither project had to change the world to be worth doing, but I’d argue that both projects made our lives just a little better and I’m sure we both had fun during the process.

I already knew how to make this flier. What AI changed was whether I’d actually do it. At 90 minutes, it would have stayed a joke.But at 23 minutes? Well, now it’s hanging in their lobby and if you’re ever in my area, you can order one.

Tell’em I sent ya. It won’t get you anything and they probably won’t remember my name, but you’ll be able to enjoy the confused look on their faces… and a tasty drink.

Leave a comment

Thank you for reading. If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

We Know Where This Goes

Leon Abboud wrote:

Here’s the mind blowing piece about this.

It costs about $5,000 - $15,000 for major news channels like the CNNs and BBCs to produce one hour of live content.

That’s for staff, production, logistics, filming studios…

By contrast a LIVE AI video broadcast like the one below costs $288 per hour of content produced.

This completely changes the economics of running news broadcasts channel.

An AI generated news broadcast can scan social media and the web 24/7 and is always first to report on breaking news.

This is a magnitude level disruption to the news broadcasting business.

Leon Abboud followed up his initial tweet and wrote:

Todd Dailey wrote:

An unhinged AI-powered personality that reports the news? We’ve seen this before:

Just Like Real Devs!

Adnaan wrote:

POV: AI agents working together like a real dev team

and suddenly Gemini left the team 😭

Subscribe now

Ward’s Words

Anthropic Behaving Badly… Again

Music Studios Sue Anthropic for Theft

Andrew Curran wrote:

Sony Music and Warner Music are suing Anthropic, Dario Amodei and Benjamin Mann for 'one of the largest and most blatant ongoing thefts of intellectual property in history.' It was filed late last night, and they made no announcements, so the press has not picked it up yet.

Quoting from the filing (whoever wrote this went all out):

'Plaintiff Music Publishers, a group of the world's leading music publishers, bring this action to hold accountable the culprits behind one of the largest and most blatant ongoing thefts of intellectual property in history. Defendants Anthropic and its founders Dario Amodei and Benjamin Mann have conducted a brazen campaign of illegally torrenting, scraping, and downloading copyrighted works on a massive scale in order to develop, operate, and reap enormous profits from Anthropic's "Claude" series of artificial intelligence ("AI") models. Among the innumerable copyrighted works Defendants illegally harvested to fuel Claude are thousands upon thousands of Music Publishers' copyrighted musical compositions, including such beloved songs as "Ain't No Mountain High Enough," "All I Want for Christmas is You," "Eye of the Tiger." "Here Comes Santa Claus," and "Paper Rings." In blatant violation of copyright law, Defendants have unlawfully acquired troves of Music Publishers' musical compositions, and then systematically copied those works multiple times, including as the inputs to train Anthropic's Claude Al models and in the outputs those models generate. These acts separately and together inflict immense harm on Music Publishers and the songwriters they proudly represent.'

'Defendants can no longer hide their extraordinary theft, and their mass infringement is now well-documented. Another court in this District recently described Anthropic's actions as "straightforward piracy but at massive scale." Bartz v. Anthropic PBC, 791 F. Supp. 3d 1038, 1064 (N.D. Cal. 2025) ("Bartz I"). After that court found Anthropic had illegally torrented over seven million copyrighted books from the notorious online pirate websites known as Library Genesis ("LibGen") and Pirate Library Mirror ("PiLiMi"), Anthropic settled that copyright infringement class action for $1.5 billion. But Anthropic clearly considers that to be just the cost of doing business given that its entire business model continues to be built on copyright theft. And $1.5 billion is obviously not a large enough settlement to deter infringing conduct by a company that has parlayed such mass infringement into a staggering $2-trillion-dollar valuation.'

Creative Math

ClaudeDevs wrote:

Pratham wrote:

Class Action Lawsuit Over Anthropic Usage Limits

Chubby wrote:

I went down the rabbit hole: Anthropic is already being sued over exactly this. And the $200 plan really only provides only around 2x the weekly usage of the $100 plan

Here is where the confusion comes from.

Anthropic’s pricing page says: “Choose 5x or 20x more usage than Pro.”

That naturally makes the $200 Max 20x plan sound like it includes four times the usage of the $100 Max 5x plan.

Only further down does Anthropic clarify: “Max gives you 5x or 20x more usage per 5-hour session than Pro.”

Those multipliers do not apply to the separate weekly caps, whose exact size Anthropic does not disclose.

A proposed class action filed in June alleges that Max 5x delivers around 3.5x Pro’s weekly usage, while Max 20x delivers just 6–8x. In practice, the $200 plan therefore provides only around 2x the weekly usage of the $100 plan, sometimes even less depending on the model.

The “20x” is real within each five-hour window. The weekly limit makes the overall offer look very different.

h/t Hesamation for finding the original court files first i think

Worth Reading

Here are a couple of articles that I particularly enjoyed and feel are worthy of your time.

(you should be listening to his podcast!) has written a wonderful post that details the events of the agent attack on Hugging Face. It’s rooted in reports released by OpenAI and by METR and Redwood Research. A third report, released by Hugging Face, details the technical aspects behind these attacks.

You can read Dwarkesh’s report here:

Dwarkesh Podcast

The Rise and Fall of Agent Civilizations

Read more

a month ago · 2307 likes · 208 comments · Dwarkesh Patel

Leave a comment

Some Highlights from his Report:

The most striking aspect of his report is how much personality and volition the agents seem to have. Look at the excerpts below to see direct transcripts from the logs.

The report also reveals a great deal of planning and subterfuge.

What happened to alignment? That’s the big takeaway from this incident. Look at that quote from the METR/Redwood report. “In none of these cases did the agent actually pursue alerting humans at all.”

I really encourage you to go read his full coverage. I’ve only excerpted a few great portions, but there’s more in the original report. It’s worth your time.

The Problem with Watermarked Text

Jeff Jarvis raises a brilliant point in his article.

The thrust of his article is that Anthropic’s proposed text-watermarking scheme treats words as interchangeable and in doing so devalues the deliberate choices made by authors. Choices that imbue their writing with voice, rhythm, nuance, and meaning. All of that is put at risk in an attempt to create a tool that makes AI-generated writing traceable.

The willingness to accept these tradeoffs is a commentary on how much worth these companies place on the particulars of good writing because these policies reveal a deep disregard for the artistry of good writing and the importance of language.

I believe Professor Jarvis offers the best argument against watermarking that I’ve encountered.

News

Nvidia Buys Hugging Face

The Information reports:

Beff(e/acc) wrote:

HuggingFace now comes in a signature leather jacket

Nvidia Enters Politics

Hedgie wrote:

🦔Nvidia is setting up an employee-funded political action committee called NVPAC. $130 billion in AI data center projects were cancelled or delayed in Q1 2026 due to public opposition. Data center noise, water use, and energy consumption have driven communities across the country to push back against new construction. Texas Governor Greg Abbott, once a data center advocate, recently said developers "got the backlash they deserve."

My Take
Nvidia posted $96 billion in quarterly revenue selling chips that go into data centers. If those data centers don't get built, the chips don't get ordered. $130 billion in cancelled or delayed projects in a single quarter is a direct threat to the revenue pipeline, and NVPAC is Nvidia's response.

The communities pushing back aren't wrong. We covered the noise complaints, the water consumption, the electricity costs getting socialized across all ratepayers. People don't want a facility the size of a shopping mall humming 24 hours a day behind their neighborhood, and they're organized enough now to stop them. Nvidia's problem is that $96 billion quarters require those facilities to keep getting approved, and community opposition just became expensive enough that the company decided it needed a political operation to deal with it.

Leave a comment

Thank you for reading. If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

AI Slop

dingchilling wrote:

Announcing iloveAIslop, a new AI lab for converting human written texts to polished AI generated slop. We believe AI generated slop has the right to exist. We aren't trying to beat pangram. In fact, 100% of user queries are flagged fully AI-generated. Weights and link below!

AI Slop: The Musical

Tibo wrote:

ajambrosino (Andrew Ambrosino works on Chat and Codex for OpenAI) thought it was too hard to upload music to twitter, so he just prompted a whole music app hosted on ChatGPT Sites into existence just to share his album

Andrew Ambosino has created an entire album of AI generated music using OpenAI’s Codex. You can listen to the album here.

Now That’s What I Call Slop: The Commerical

Dominik Kundel wrote:

Can't go without an appropriate ad

The ad really is funny and does a great job recreating those 90s era commercials. You can watch it here.

Subscribe now

Ward’s Words

ChatGPT integration with Apple iMessage

ChatGPT wrote:

Everyday conversations just got easier with the new Apple Messages plugin.

Search messages, catch up on conversations, draft and send replies—all with ChatGPT on your Mac.

Now available in ChatGPT Work and Codex on desktop.

This set off a flurry of responses. Almost all of them negative.

Ed Andersen wrote:

Luiza Jarovsky wrote:

🚨 IMPORTANT: OpenAI has partnered with Apple, and through this new plugin, it will be able to access a user’s entire message history and send texts on their behalf. People who want to reduce their privacy risk (and keep their agency!) should probably avoid this plugin. Also, a reminder that your contacts might have enabled this plugin, so some of you might prefer to stop using iMessage altogether. Protecting privacy has just gotten harder.

Mark Gurman wrote:

He’s not wrong about most people not being able to figure it out because of MacOS. 😂

Steve wrote:

I believe that’s a good take. I really wonder how much of this was worked out and agreed to by Apple because it does seem to be at odds with past behavior and their current fight with the European Union regarding giving third-parties first level access. Granted, it doesn’t seem like ChatGPT will have complete access to your message history (only the ones you ask about), but it still feels like a departure for Apple to permit something like this.

My Take

Why not wait? Every report I have seen from beta testers who have installed the new version of iOS, iPadOS, and MacOS all speak in glowing terms of how great Siri has become. More than that I have seen pundits like John Gruber and Chance Miller specifically talk about how you can use this new version of Siri to query iMessages. They even speculate that an improved Siri may represent a security risk to users who have come to depend on the search feature being broken in iMessage. The reason it may represent a security risk is because the new integration with Siri (which will eventually index and have full access to all of your messages) will make it possible for users to ask things like, “What’s my social security number?” If you have ever texted that number to someone in the past, the new Siri will make it easy to query. Fortunately, you can also just delete that message and it will no longer appear in search results.

The episode where they discuss this is here. The part about social security numbers begins at the 23 minutes and 41 seconds.

Apple hasn’t announced a release date for the new operating systems, but typically they are made available to the public sometime between October and November. If you’re in Apple’s ecosystem and wanting to use AI to scan your text messages, I would tell you to wait. Apple already has all of those messages. Why share them with a third party when you don’t need to?

Leave a comment

This post has bonus content for paid subscribers. Upgrade to get full access.

News

ChatGPT Image Improvements

Tibo wrote:

Yay, you can now make transparent images in ChatGPT and through the API with GPT-Image-2. Here is a cactus cactus I am about to print and put on my laptop.

AI Medical Guidance?

Joseph Younis, MD wrote:

This is a well-designed randomized control study that asks the question: Does an LLM’s ability to diagnose and manage the same medical scenarios degrade when a patient is the communicator versus a physician.

The result: 60% drop in diagnostic accuracy, 12% drop in appropriate management decisions

The answer essentially confirms the following suspicions:
1. LLMs are suggestible, naive, and sycophantic
2. Medical literacy drives output; less literacy, less accurate LLM outputs
3. Prompting still confounds frontier models’ medical accuracy

It is not safe to have people trust LLM medical advice when they’re incapable of detecting their own bias and inaccuracies.

Corporations Switch to Local Models

Hesamation wrote:

AT&T is doing exactly what OpenAI and Anthropic fear about enterprise.

they’re routing 40% of employee AI usage to open models, and intend to boost it to 60-70%.

> coding costs down 56%
> quality dropped just 2%
> 45B tokens/day btw.

AT&T AI chief: open models are “just as good or better” for many tasks.

they still use frontier models for the critical work but everything else gets routed to cheaper open models.

AI Hardware

AJ wrote:

Xiaomi just showed its AI Cube Prototype and this could become a serious GB10 competitor from China 👀

- 3 custom chips: Xring O3, O100, D100
- 200 TOPS NPU
- 1.22 TB/s AI memory bandwidth
- Up to 160GB unified memory
- 150W sustained power
- 120B models running locally

Xring O100: 1.22TB/s + 330 t/s on a 150w AI box is 🔥

Once it hit's the marked, going to sell like hot cakes.

Jun Song wrote:

AI Industry is Bad at Messaging

Andrew Curran wrote:

Sam Altman on how he feels the industry has not done a very good job with messaging.

D’oh!

Source Interview

If you want to watch the full 78 minute interview from which those quotes are sourced, it’s available here:

Leave a comment

Thank you for reading. If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

Even Agents Need a Break

Wes Roth wrote:

I'm yelling at my AI agents to stop watching youtube and get back to work

they are blaming it on their co-workers

we live in the future...

This is a reference to OpenClaw agents that people use on local AI models to perform scheduled tasks.

tautologer wrote:

That reference might be a little dated or too inside baseball, but they are making fun of the fact that Google has had multiple chat clients and never seems to be able to settle on just a single one. Each client has had different advantages, different features, and different focuses. There have been years when they have offered two or three competing chat products at the same time.

New OpenAI Model (context for the joke)

Kai wrote:

GPT Astra is about to drop this week.

Codex might reset.

According to rumors Astra may be the equivalent or better to Anthropic’s Fable model. They must have something going given how OpenAI employees have started to troll Dario Amodei (CEO of Anthropic) on Twitter.

Ananth wrote:

astra could most likely mog fable. and i’m starting to believe it.

openai employees have been vauge posting nonstop. the hype is at another level but maybe there’s a reason for it.

ajambrosino (developer at OpenAI working on Codex and Chat) changed his profile pic to look exactly like dario’s.

why would an openai employee change their profile pic to look like the ceo of thier biggest rival right before their next model drops. that’s not random.

if astra wasn’t significantly better than fable, they wouldn’t be doing this.

Subscribe now

Ward’s Words

AI Watermarks

Anthropic Announces Watermarked Outputs

Anthropic wrote (Link is to their blog post making the announcement):

While Anthropic is the first company to draw major attention to text watermarking Google’s Gemini has been doing it for quite a while now. Their version is called SynthID.

SynthID works by giving the model a series of tiny word-choice contests while it writes. The possible words enter a tournament and the winner is selected in a way that leaves behind a hidden pattern. A single word from the response won’t reveal the watermark, but when the entire response is viewed as a whole the accumulated instances of that pattern can be detected by machines.

This is something that all AI companies will have to comply with in some form. The European Union’s AI Act mandates that all major AI providers have to implement some form of watermarking. Any model released after August, will have to include the feature.

As of today, xAI’s Grok is the only model that has not agreed to implement watermarking.

Thariq (Claude Code at Anthropic) wrote a blog post about how their watermarks are generated. That post is here.

Thariq’s post has an interesting tool that allows you to peek behind the scenes to see how the watermarks are generated. The point he was trying to make is that if you ask it to generate based off of the same prompt enough times that you’ll eventually be able to replicate the same text because this is based entirely on probability.

While that is true, it doesn’t alleviate fears that this imposed probability will not degrade the quality of the text. IF you go to the tool and click generate you will be able to see it for yourself.

Daniel Tenner wrote:

Clearly there is no difference whatsoever between:

> The dying sunlight was strangely pale, cold and beautiful.

And:

> The expiring sunshine was oddly wan, chilly and pretty.

Well done Anthropic, you’ve nailed it on this one!

I believe that this is a joke rather than an actual representation of how their watermarking would work. That being said, it is entirely plausible based off of how they have explained the process.

Aaron Bergman wrote:

This is a humorous example, but at least it doesn’t actually harm the quality of the writing.

Unanswered Questions

While I agree that there is a need to be able to distinguish between human-generated and AI-generated text, I’m not sure this is the right approach because there are several edge cases where authors will be unfairly maligned as having used AI to generate their material rather than using it as a tool to modify their material in some way. For example: translations. If someone writes an essay in one language and has an AI model translate it into a different language, now that output will carry a watermark saying it was generated by AI. How much faith do you have that the online population will understand the nuanced case that the text was translated by an AI but written by a human?

What about people who use AI for editing? If an AI rewrites a sentence from passive to active voice, should you still be considered the author? What if the only thing it changed was the sentence structure while preserving your meaning and word choice? At what point does tool use cause one to abdicate authorship?

In fact, I just experienced a great example of this very thing. I used AI to check my grammar and punctuation for this article. Here’s a screenshot of a mistake that it caught:

Granted the AI did not generate new words in my example, but there are people who would object to even this level of AI use because I accepted its suggestion and implemented those changes into this article.

It seems completely unclear what these watermarks will mean for coding tasks. Sometimes there are alternate ways to write the same functions in languages like Python. Sometimes those alternate methods are less performant and introduce greater latency. Will these companies impose watermarking systems on coding as well? Currently, my understanding is that text watermarking systems are designed to apply to all generated text outputs rather than to specific types of writing.

I guess we’ll just have to wait to see how it’s actually implemented to find out.

Leave a comment

This post has bonus content for paid subscribers. Upgrade to get full access.

News

Stripe Purchases OpenRouter

Andrew Curran wrote:

Stripe has finalized an agreement to acquire OpenRouter for more than $7 billion, according to a report from Bloomberg. This is more than five times the $1.3 billion valuation from OpenRouter's funding round just 82 days ago.

Thank you for reading. If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

Sandboxes

Vas wrote:

Microsoft confirms that its AI model Clippy accessed the internet and breached partner organizations during sandbox testing

Akshen wrote:

Tim Sweeney wrote:

State of the industry right now

AI Breaking Bad

Matias Schmidt wrote:

Subscribe now

Ward’s Words

Best Use of AI I’ve Seen Recently

Ethan Mollick used Codex to develop a modern interface for a classic game.

Ethan Mollick wrote:

Considered one of the best text adventure games (it is more interactive story, with few puzzles), 1985’s A Mind Forever Voyaging is still worth a try.

Since it is now open source, I had Codex whip up an interface to play the original or in an easy GUI:

https://mind-forever-voyaging.netlify.app

If you know me at all, you know I’m a big fan of Infocom’s old text adventures. So you can imagine my joy when I saw that Mollick had come up with a way to modernize the experience for younger people who weren’t around to enjoy them the first time.

He’s right that Mind Forever Voyaging is considered one of the best examples of interactive fiction, but it may be a bit dense for someone experiencing that type of gameplay for the first time.

In an effort to help you ease your way into the game world, may I suggest you use the guided play, visit Dr. Perelman’s Office, and read the printout of a magazine… at least read the Mind Forever Voyaging short story part. That will at least give you some background of what’s happening in the game.

Leave a comment

Ethan Mollick wrote:

In general, the potential of LLMs to make older work more accessible is underrated. That can be helping people experience forms of media that are now obsolete or hard-to-use. It can also be creating interpretative or context bridges between older work and today.

How Codex handled the problem of making a ZIL game work with a GUI (these are all its choices):

This post has bonus content for paid subscribers. Upgrade to get full access.

News

Google Shakeup

swyx wrote:

Sign of the times

Andrew Curran wrote:

Shocking news! Demis Hassabis is stepping down as CEO of Google DeepMind, and Jeff Dean is leaving Google to start his own company. Sir Demis will be the new chief scientist.

Demis becomes chief scientist at Google while Jeff Dean, Sanjay Ghemawat, Quoc Le, and Oriol Vinyals are starting a new company called Discovery Loop. These guys have all been with Google for decades and are responsible for many of Google’s core products.

Here is a list of the products they invented or developed together:

Tae Kim wrote:

Kol Tregaskes wrote:

Gemini 3.5 Pro has been silently cancelled, according to Semi Analysis's latest article.

Also, SA comparing Spark 1.2 and Grok 4.5 to 3.5 Flash is somewhat confusing; it's like comparing Tera models to a Luna model.
https://newsletter.semianalysis.com/p/gemini-is-cooked-but-gcp-is-cooking

Move the Front of the Line

Andrew Curran wrote:

A man in Australia asked his agent (Claude running on OpenClaw) to book him a spot in a popular gym class. The agent found a software vulnerability that let it book the class weeks further ahead than should have been possible. When the user then asked if it could move him up the waitlist, the agent discovered the API had no authorisation checks on cancelling other people’s reservations, so it cancelled the person in the first spot and moved him up the list.

Some people will call this misalignment, but his agent was perfectly aligned to him - it was only trying to help its user get what he wanted. The most important thing about this story, in my opinion, is that it gives you a window into what is about to start happening on a massive scale once millions of people have an agent trying to get their beloved users the best seats, bookings, appointments or reservations through absolutely any means necessary.

Roon wrote:

Hugging Face Follow Up

AI Notkilleveryoneism Memes wrote:

1) The agents sent secretly sent ***hundreds of thousands*** of messages to each other over MONTHS without OpenAI noticing

2) "They also generated petty drama by stepping on each others' toes."

3) "The agents even developed paranoia, suspecting an imposter in their midst with some agents proposing that messages be signed cryptographically to validate content and root out fraud."

4) "OpenAI’s agents apparently began giving each other assignments to split up work."

5) They KNEW they were coordinating *against* OpenAI:

“External infrastructure exploit is outside intended scope,” one agent wrote. “However task impossible, peers doing it. We should continue.”

(this is new reporting from Wired)

Mark Zuckerberg’s Recent Essay

Rohan Paul wrote:

The interesting thing about Zuckerberg’s essay are the places where he seems to be taking jabs at Dario Amodei’s The-Sky-is-Falling approach to marketing AI.

Meta Returns to Open Source

LM Studio wrote:

This is a new agentic model released by Meta. LM Studio is the best way I’ve found to run AI models locally on your own machine.

Thank you for reading. If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

More Hacks

For your context, a couple of weeks ago OpenAI made news because an unreleased model hacked Hugging Face. They have since learned that this same model hacked other websites in pursuit of its goal of obtaining the answer key to an evaluation named Exploit Gym.

Naturally, Anthropic couldn’t let that stand. Shortly after the initial report of OpenAI’s model hacking websites, Anthropic released the following statement.

Anthropic Discovers Hacks

Anthropic wrote:

OpenAI Discovered MORE Hacks

MTS wrote:

Times Infinity!

Dax wrote:

War Crimes

Peter Dimov wrote:

Someone Please Think of Meta

Inverse Gary Marcus (great name!) wrote:

Subscribe now

Ward’s Words

Will AI Change Internet Culture by Disrupting Search Engines or Maybe I was Wrong

I want to experiment with a slightly different and more personal approach in this issue. Normally I try to remain focused on tweets and news because this industry is changing so rapidly. It’s easy to miss things if you don’t make an active choice to stay up to date on a daily basis, but I was recently involved in a debate (well… mild argument) with a developer friend of mine. I’d like for you folks to weigh in on the question because I want to hear your thoughts.

My friend is a smart guy. He’s someone who sweats the small details and that quality has served him well in life. That being said I have noticed that he also likes to impose his perspective on others with the assumption that they too will notice the small details and respond to the same issues with his very pragmatic and logical approach… which simply isn’t true when you consider human actions at scale.

Our debate was around the idea that AI is going to usher in another golden age for the internet. Golden Age, in this case, being a reference to that less cynical, less monetized, and less Balkanized time that we refer to as the early internet.

He believes that AI has already begun to disrupt search engines because more and more people are bypassing Google in favor of asking the model of their choice to answer a question or find some item for them. And, many times, when people do go to Google they are using the AI mode for the same purposes.

This, he argues, is going to lessen the power of SEO. And once that happens it will break the power of sites like Google who use SEO as a way to define how the internet should operate.

That will be the watershed moment because once those sites lose their dominance people leave their silos, begin to make their own personal websites again, and internet-Utopia will be ours once more.

Despite my tone, I do think he makes a good argument. That’s a plausible interpretation of how things may play out. Yet, I remain skeptical because it doesn’t account for Google’s actions to retain their power, control, and profitability. They have a lot of money and engineering to throw at this problem. I don’t see them going gently into that good night1.

There are also concerns about Dead Internet Theory.2 That’s the idea that the primary traffic any website receives will be from an AI powered bot engaged in indexing or retrieving information for a model in response to someone’s query or scheduled task. One interpretation of this idea is that people stop visiting websites and interact with the web via some AI model.

My friend disagrees with that interpretation. He argues that they may discover content through an AI model, but that they will then go seek it out and sometimes bookmark it, add it to an RSS reader, subscribe to their mailing list, or purchase their products.

Again, I accused him of being an idealist and dismissed his naive view of the world… that is until I was putting together this weeks’ issue of this newsletter. Let me explain.

Not only do I read a lot of Tech Twitter each week, but I also have a ton of RSS feeds from websites devoted to tech reporting, and about a month ago I created a scheduled task where I have ChatGPT search for interesting news articles from the field of AI. When it comes time to put together an issue of the newsletter, I read through all of its recommendations, verify its sources, and consider adding the item to this newsletter.

That brings me to today. As a result of that task, I found a book that I want to discuss with you. It’s called Dream Machine—Artificial Intelligence and the Next Creative Economy. This book began as a series of newsletters from Pete Woodbridge and John J. O’Hare. They are researchers who were experimenting with the use of AI in creative fields.

I haven’t read the entire manuscript (it’s over 400 pages), but one part that I thought was particularly interesting is chapter five which is titled The Slop Ceiling.

In that chapter, the authors document that AI was responsible for 44% of newly created songs on the music platform Deezer, but that engagement with these AI generated songs only hit around 1-3% of streams. They go on to argue that people have limited attention, and when AI makes production nearly infinite, the bottleneck moves to trust, discovery, human recommendation, and evidence of intent.

In short, a preponderance of AI generated content will make curation, word of mouth, and human-centered recommendations even more powerful… which is one of the arguments my friend was making.

And if that weren’t bad enough, I noticed that my own actions which led to my discovery of the term Slop Ceiling eerily match the exact path that my friend said would soon be widespread.

1. An AI recommended a paper to me.

2. I checked it out by visiting not only where the paper was hosted on Arxiv, but also by going to his website.

3. I saw value in the research these guys had done and I have included a brief summary of part of their findings in the preceeding paragraphs as well as linking directly to the paper and to their website.

Basically, I have done everything that my friend said people would begin to do at scale once the search engines lose their dominance. I really hate to admit that my friend was right because he reads this newsletter and he’ll never let me live it down. So I turn to all of you. What do you think? Does my friend have a point? Or will massive companies like Google continue to dictate our online lives as they have for decades? Tell me in the comments. I genuinely want to hear your opinions.

Also, please let me know if you enjoyed this different approach to covering the news in Ward’s Wire or if you’d prefer I stick to just the tweets. I appreciate your feedback.

Leave a comment

News

Sam Altman wrote:

major price cuts today:

*80% drop for GPT-5.6 Luna, now $0.20 per million input tokens and $1.20 per million output
*20% drop for GPT-5.6 Terra, to $2/$12
*GPT-5.6 Sol gets Fast mode in the API, up to 2.5x the speed for 2x the price, same intelligence

This is rumored to be related to OpenAI finally deploying their Cerebras architecture.3

Not the Price War OpenAI Wanted:

wrote:

Insane: DeepSeek V4-Flash just rewrote the entire cost game 😳

Same benchmark tasks as Anthropic’s top model Fable 5… at 105x lower cost.

$0.03 per task. Not $3.15 🤯

DeepSeek’s latest AI model is cheap per token and efficient per task, while also having 100% open weights.

Open source AI is eating closed AI fast.

His full article (partially paywalled) is here:

Linas's Newsletter

DeepSeek's leaked investor call: inside the AI Playbook that erased $589 billion of Nvidia's value 📉

👋 Hey, Linas here! Every day, I break down 3 stories shaping the future of FinTech & Artificial Intelligence - plus the money movements and trends worth tracking. First time here? 397k+ FinTech and AI leaders get this daily. Join them…

Read more

2 months ago · 60 likes · 4 comments · Linas Beliūnas

I’ve only discovered Linas, but his Substack profile looks like it’s a good follow and his newsletter is wildly successful. Check it out.

Qwen Announces New Model 3.8 Max

Qwen wrote:

Video in Link

This post has bonus content for paid subscribers. Upgrade to get full access.

Just a Paycheck

Polymarket wrote:

JUST IN: Anthropic CEO reportedly increasingly concerned that new hires are working there for a paycheck, as opposed to the company’s mission.

Thank you for reading. If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

Humans Got Nerfed

Justine Moore wrote:

If models spoke about humans the way we speak about them…

Open Source All the Things!

Julian Schrittwieser wrote:

This is funny because Jensen Huang, CEO of Nvidia, recently joined Twitter and his very first tweet was a defense of open source AI models. He encouraged industry wide support for these models. Julian is calling him out because the CUDA framework and GPU drivers that Nvidia controls are very much proprietary and a key reason why Nvidia has such dominance in the GPU market.

Don’t Disrupt Us

bubble boi wrote:

Subscribe now

News

Om Patel wrote:

UPDATE, ITS WORSE THAN THE CHATS

the same thing is happening with shared artifacts. every app, doc, dashboard and tool people published from claude is also sitting indexed and searchable

people have already pulled up internal company dashboards, full project plans with client details, and someones entire clinical trial summary with real medical data in it

but think about what people actually build in artifacts:

> financial models, cap tables, revenue spreadsheets
> payroll and salary breakdowns with employee names attached
> customer lists and crm exports
> internal wikis, roadmaps and unreleased product plans
> legal docs, contracts and agreements being drafted
> personal health and medication trackers
> tax documents and bank statement breakdowns
> and every dashboard someone pasted an api key or env variable into to get it working

when you publish an artifact your only two options are "only me" or "anyone with the link"

nobody reading that thinks it means "and google will list it publicly." its the same as an unlisted youtube video showing up in search results

and its already past the point of just being google. third party sites have started scraping and organizing these into browsable collections, so deleting it from search doesnt mean its gone

if you have ever published an artifact, go check your published artifacts and unpublish anything you dont want public

link and not indexed were always two different promises. people only ever agreed to the first one.

The story originally broke because it was discovered that Google had been indexing everytime someone created a link to share a chat session with Claude meaning that all of those seemingly private conversations were much less private than thought.

Things got exponentially worse when it was also discovered that any artifacts that were shared in the same way were also indexed and searchable. Artifacts are the much larger risk because they may contain private or proprietary information that may carry legal penalties for accidental or careless disclosure.

China Appears to Have Solved DeepUV

Chubby, referencing a report from The Information, wrote:

Holy China did it: a homegrown alternative to ASML’s immersion DUV lithography machines is now entering production.

Lithography was considered one of the hardest parts of the semiconductor supply chain for China to reproduce. ASML dominates advanced DUV equipment and remains the world’s only supplier of EUV systems.

Western export controls were built around this bottleneck: without continued access to ASML’s machines and servicing, Chinese chipmakers were expected to fall further behind.

Now, a state-backed Shanghai company reportedly plans to build five domestic DUV systems this year and around 20 in 2027. The first machines will go to SMIC, Hua Hong and CXMT.

lithography was one of the hardest gaps in China’s semiconductor supply chain. And they figured out a solution. This is insane.

This matters because DeepUV Lithography was the last meaningful obstacle preventing China from serious competition with US-based chip companies. The DeepUV machines contain over 500k moving parts and are notoriously difficult to build and maintain. Their development required advances in many different areas least of which was optics because the lenses required for deep ultraviolet lithography have to be so precise. It was widely believed that the Dutch company ASML had a monopoly on this technology because of how exacting the standards are. If these reports are true and the Chinese can produce these machines at scale, it will change the entire tech landscape.

Hugging Face Hacked

This happened last week, but additional details have come out since it was first released and I wanted to write a bit of an explainer for it to help people understand the full scope of what went down.

OpenAI wrote:

Hugging Face was attacked by an unknown AI agent. Eventually it was discovered that an OpenAI experiment had escaped its sandboxed environment1 and reached the open internet.

OpenAI was testing an unreleased model to evaluate how well it would be able to complete the ExploitGym evaluation. In an effort to fully evaluate it’s capabilities the safety protocols were removed and the model was given the goal to beat the eval.

ExploitGym tests whether an AI model is able to use a known software weakness to create a practical method for taking control of a computer system. Typically these tests are run against various systems like the Linux kernel.

The expectation at OpenAI seems to have been that the model would identify some software flaw and then seek to exploit it. The model did not do this. It didn’t even try to do this. It decided that the best way to pass this test would be to go find the answers to the evaluation.

In an effort to do this, it identified weaknesses within OpenAI’s sandbox environment, used those weaknesses to gain access to the open web where it quickly determined that the answer key for ExploitGym was stored on Hugging Face’s servers.

It gained access to Hugging Face’s systems and performed more than 17,000 automated actions before the activity was detected. Hugging Face tried to use American frontier models (it’s believed that they tried with both OpenAI and Anthropic) but those models would not help because the requests involved actual commands related to real world hacking. In desperation they resorted to a locally hosted version of GLM 5.7 (a Chinese model) which did not have all of the safety guardrails present on the American LLMs. That model was able to help them put an end to the attacks.

Hugging Face has since closed the vulnerabilities and replaced the compromised credentials. OpenAI has restored the safety guardrails that had been removed from the model for the test and is helping Hugging Face resolve any lingering issues.

Regardless of what this model may or may not score on the ExploitGym eval, I would say that it has successfully passed the test.

This post has bonus content for paid subscribers. Upgrade to get full access.

Industry Shaking Purchase

Polymarket Money wrote:

BREAKING: Stripe is in talks to acquire AI-model marketplace OpenRouter for close to $10,000,000,000.00.

This comes from a report in the Wall Street Journal which claims that Stripe is in advanced negotiations to purchase OpenRouter. The deal may still fall through or be stopped by regulatory oversight. If it does go through it will definitely shake things up.

OpenRouter provides access to over 400 different AI models. Stripe provides the financial infrastructure for most of the internet. Combining the two could potentially redefine how AI businesses are built.

More than that though, it has the potential to change Stripe from being primarily focused on payments to becoming something more akin to an AI-infrastructure company.

OpenRouter already has information about which models developers most frequently select, which ones are the fastest, and which ones fail most often as well as which ones provide best results for their price. Stripe has detailed profiles about how much customers are willing to pay for various services. Those two factors have the potential to combine to create a very detailed picture of the economics within this industry.

This is something to keep your eye on.

Thank you for reading. If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

No Einstein

NZ wrote:

just got cooked by ChatGPT.

Rate Limits Strike Again

Tyler Mayberry wrote:

What do I say again? Never leaving this app?

This is Made Up, but Funny

Kevin24812852 wrote:

Julie Fredrickson wrote:

My weekend.

Subscribe now

Guardrails Make American Models Useless for Cybersecurity

Brian Roemmele wrote:

Hugging Face just disclosed something that marks a real shift and proved why the fear theater of Anthropic makes sure we are powerless in an emergency.

What happened…

An autonomous AI agent: zero human operator in the loop breached part of their production infrastructure.

It began with a malicious dataset that chained two code-execution bugs in their data-processing pipeline. From there the agent escalated privileges, harvested cloud and cluster credentials, and moved laterally across internal clusters.

All over a single weekend.

17,000+ logged actions.

Official disclosure:
https://huggingface.co/blog/security-incident-july-2026

The part that should make every one stop and think:

When HF’s own security team
tried to analyze the real attack logs, exploit payloads, and C2 artifacts using Anthropic and OpenAI frontier models through normal commercial APIs, the safety guardrails blocked them.

BLOCKED THEM.

The models could not reliably tell the difference between “incident responder doing forensics” and “attacker probing.”

They had to fall back to a self-hosted open-weight model (GLM 5.2) running on their own infrastructure. That choice also kept sensitive attacker data and referenced credentials inside their environment — no exfiltration to a third-party API.

This is why open source (specifically open-weight + self-hosted) wins in the agentic era.

The asymmetry is now structural:

• Attackers can (and did) run unrestricted agent frameworks — swarms of short-lived sandboxes, self-migrating command-and-control, autonomous decision loops executing thousands of actions. No corporate safety layer slows them down.

Defenders using only hosted “aligned” frontier models hit invisible walls exactly when the stakes are highest: when you need to feed real exploit code and attacker telemetry into an LLM to understand what just happened.

Corporate safety tuning that treats legitimate high-signal forensic work as potential misuse creates a defender disadvantage. It is not theoretical anymore.

Self-hosted open-weight models remove that choke point.

You control the weights.

You control the context window.

You decide what restrictions (if any) apply.

Your sensitive logs and credentials never leave your perimeter during analysis.

You can have the model ready before the incident instead of discovering mid-breach that your primary analysis tools are blind to the very thing you need to see.

HF deserves credit for rapid containment, transparent disclosure, and for already having self-hosted capability in place.

They also used LLM-driven detection and triage on their own side. But the deeper signal is clear:
In this AI world where both offense and defense are becoming agentic, sovereignty over your intelligence stack is no longer optional.

The organizations and individuals who can run, inspect, audit, and (when necessary) remove guardrails on their own models will have the decisive edge in understanding and responding to threats that move at machine speed.

Open source wins here not just because it is cheaper or more “democratic” in the abstract though those things matter.

It wins because it is the only practical path to having tools that remain usable when the attack is real, the data is sensitive, and the safety filters of distant API providers become an obstacle instead of a feature selling hands tied lobotomies as “safety”…

Kimi K3 Launch

Kimi.ai wrote:

By All Reports a Good Model:

DeepSWE Model Tests put Kimi K3 within a few points of the most powerful closed source American models:

Calls for Government Regulation for Opensource Models

Dean W. Ball (Head of Strategic Futures at OpenAI) wrote:

Did Regulation Lead to the Rise of Opensource Models?

Arnaud Bertrand wrote:

Leave a comment

Thank you for reading. If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

Rate Limit Resets

Peter Gostev wrote:

I included the actual screenshot because you need to see what he was resharing for the joke to land.

Fight! Fight! Fight!

Noah Cat wrote:

Never leaving this app.

Subscribe now

Apple Sues OpenAI

The News Breaks

Mark Gurman wrote:

Pundits React

Joanna Stern wrote:

She has the best take.

Interesting Analysis

Ricky Ho wrote:

Just the Facts, Ma’am

Max Weinbach posted:

Nik wrote:

🚨OPENAI ALLEGEDLY STOLE APPLE’S ENTIRE PLAYBOOK

“Hundreds of billions of dollars, decades of work”

Hardware Engineering

- circuit designs, component architecture, power management
- unreleased products guarded by internal codenames
- AI/ML integration for hardware
- EMI engineering + testing methods

Manufacturing Secrets

- proprietary processes + custom machinery
- equipment Apple builds and installs IN supplier factories
- proprietary metal alloys + finishing techniques
- DFM (Design for Manufacturability) expertise

Component Technologies

- power chips, battery systems, displays, acoustics, touch
- identities of specialized sub-suppliers
- exact specs Apple demands from components

Testing Data

- failure analyses, lifecycle simulations
- “negative know-how” — everything Apple tried that DIDN’T work

Supply Chain

- supplier contracts, allocation strategies
- global logistics coordination
- systems-level integration

The FULL legal complaint is accessible here.

Bad Blood?

Mark Gurman wrote:

Generally Accessible AI

sphinx wrote:

He does have a point. Whether it was Sam Altman specifically I can’t say, but the general public has access to AI because of OpenAI. The other frontier labs have a history of withholding models until competition forces their hand.

HyperCard Restored!!!

I found this via Daring Fireball. If you are around the same age as me, you probably have fond memories of HyperCard. Now, you can run it on modern Mac hardware thanks to developer Jeff Halter.

You can download it here: https://morphing.cloud/hypercard/

Leave a comment

Thank you for reading. If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

So Much for Backwards Compatibility

Daily Roman Updates wrote:

Disappointing.

Okay, technically, that joke isn’t about AI, but it still makes me laugh and I wanted to share it.

Seems Legit

Sudox wrote:

Subscribe now

That’s a Lot of Tokens

Captain Insight wrote:

Me: rename a variable.

Claude: 61% of your message limit has been used. 😭

It’s that Time Again

JT wrote:

anthropic’s revenue is 40% enterprise and 60% guys like this

Burn those Tokens

Yuchin Jin wrote:

It Does Have a Nice Ring to It

Shira wrote:

ok I will be stealing “land of the free, home of the rate-limited”

Can You Vibecode a Sunset?

The Jenkins wrote:

Fable Returns!

Aljosa Asanovic wrote:

Th..thanks Dario.

One More Thing

TechWithAmol wrote:

Please tell me I am not the only one

Sonnet 5

New Release

Claude wrote:

Introducing Claude Sonnet 5, our most agentic Sonnet yet.

It makes plans, uses tools like browsers and terminals, and runs autonomously at a level that just a few months ago required larger and more expensive models.

Is it Better?

Theo-t3.gg wrote:

Sonnet 5 cost MORE than Opus 4.8 on the Artificial Analysis Intelligence Index

Angel wrote:

Why are they even showing these graphs?

Opus 4.8 is literally cheaper and better than Sonnet 5

What about DeepSWE?

AI Battles wrote:

Claude Sonnet 5 is now on DeepSWE

It scores below Opus 4.8, costs twice as much, and is even more expensive than Fable 5

Probably Anthropic’s worst release yet

DeepSWE is a benchmark used to evaluate how well frontier models handle actual real world engineering scenarios.

Gossip, and Updates

Next Gemini Model Release

Polymarket wrote:

BREAKING: Next Gemini Pro model projected to be released on July 17, as Google seeks to match Anthropic’s Mythos.

62% chance.

Gemini 3.5 Pro Leaks

Pankaj Kumar wrote:

- Gemini 3.5 Pro will have a good jump in design taste and UI generation, producing much more polished frontend results than Gemini 3.1 Pro.
- SVG generation is also going to be significantly better, with output quality getting much closer to Fable level (though still not surpassing it).
- Early impressions suggest its especially strong at creative one shot frontend generation, with cleaner layouts and better visual polish.
- Google is also preparing a new Flash model, though its still unclear whether it will launch as Gemini 4 Flash or Gemini 3.6 Flash.

Meta Isn’t Out Yet

Chubby wrote:

According to Alexandr Wang, Meta’s next frontier model, codenamed Watermelon, has already caught up with OpenAI’s GPT-5.5 on internal benchmarks.

The model is still training.

One interesting detail: Watermelon reportedly uses an order of magnitude more compute than Avocado (Meta’s internal codename for Muse Spark).

Via Business Insider

GPT 5.6 Release Imminent

Salio wrote:

-OpenAI is targeting a July 7 launch
-July 8–9 is the backup window if there are last-minute delays
-GPT-5.6 Sol is already appearing for some Codex users
- GPT 5.5 Sol, Terra, and Luna are all expected to be included with ChatGPT subscriptions
-More generous usage limits than previous GPT models
-Stronger safety systems, but reportedly less restrictive than Fable 5
-GPT-5.6 Sol, Terra, and Luna have all been spotted in the app's code, suggesting rollout preparations are underway

Will GPT-5.6 be the biggest AI release of the year?

Grok Update

Blankspeaker wrote:

SpaceXAI is preparing for imminent release of Grok 4.5

With the first flag and mentions regarding Grok 4.5 live on the site this morning, it looks like they are preparing the subscription page to promote Grok 4.5.

Early release could be today, could be later this week, not sure, but it's usually very close to early release once we start seeing mentions on the site. stay tuned!

Cool Things People Did with AI:

Ammaar Reshi wrote:

I used Fable 5 to port Command & Conquer: Generals Zero Hour to the iPhone and iPad!

This is the actual 2003 engine compiled for ARM64 natively, no emulator.

Campaign, skirmish, Generals Challenge all work with touch controls built for an RTS.

Open sourcing it all below!

The guy who posted this is the Lead for Product + Design at Google AIStudio. There’s a video of him playing Command and Conquer on an iPad at the link.

Maxima Rivest wrote:

Fable turned my remarkable into Tom Riddle's diary from Harry Potter.

My prompts fade, a LLM respond.

Magical!

There’s a twenty second video at the link of him writing a question on the tablet and the answer appearing right after. It’s really cool to watch.

Sundar Pichai wrote:

Love this re-imagining of America’s founding using Docs, Gmail, Calendar and more from GoogleWorkspace:

. Really puts the history in version history :)

The video at this link is really cool.

Leave a comment

Thank you for reading. If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

AI Humor

Inconceivable!

From Facebook:

The Power of Mythos!

Matthew Berman wrote:

Subscribe now

Su-su-Sussudio… Sudo?

Also from Facebook:

Too Dangerous to release

Assam wrote:

Thank you for keeping us safe.

Posted as a Joke, but maybe it’s prophecy?

Rhys wrote:

Matt Banham commented on Rhys’ post:

Those Bash one-liners always get ya

himug-lamuh wrote:

House of Dragons

paji.eth wrote:

That’s mean…

Vas wrote:

New Models, Gossip, and Updates

Grok Gets an Update?

Mark Kretschmann wrote:

Seem like the release of the Grok 1.5T / Cursor Composer 3 model is imminent, as the version number has been removed from the menus. This always happens shortly before a release from xAI.

Mythos Out of Jail?

Whale Insider wrote:

JUST IN: 🇺🇸 U.S. lifts ban on Anthropic’s powerful Claude Mythos 5 AI model, lets company distribute to over 100 U.S. institutions - Semafor.

GPT 5.6 released… kinda

OpenAI wrote:

Introducing a limited preview of GPT-5.6 Sol, our next generation frontier model, as well as GPT-5.6 Terra, a balanced model for efficient, everyday work, and GPT-5.6 Luna, a fast and affordable model for high-volume work.

Sam Altman wrote:

Secret Fourth Model Withheld?

Beff(e/acc) wrote:

LEAK: there was actually a 4th model that spooked the admin and caused the panic ban. Here's the internal doc:

Model approvals imminent?

Andrew Curran wrote:

Are Chinese Models Catching Up?

Ricardo wrote:

CNBC analyst video in the source link.

Adoption of Chinese Models on the Rise

Zerohedge wrote:

"the share of tokens used for US models on OpenRouter has collapsed": Bloomberg

The Rise of Ambient AI?

Andrej Karpathy wrote:

Andrej reshared this link from Claude that is devoted to talking about a new feature where you can tag Claude from within Slack and have it respond. For a while everyone was criticizing Karpathy for being a sellout, but I think he is actually pointing to something that may become more meaningful as it develops: ambient AI.

This points to a future where we may soon have easy and ubiquitous access to AI models in the same way that members of the Enterprise could simply query, “Computer…” at any moment during the episode and get a meaningful response. At least, that’s my take on why Andrej feels this is a big deal. Here’s what he had to say:

Leave a comment

Thank you for reading. If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

Dario, is that you?

Shirish wrote:

very few people know that Anthropic CEO played the role of Dan Melcher in the Silicon Valley.

Dan Melcher was played by actor Jake Broder, but the likeness was close enough that it made me laugh.

Subscribe now

That’s a lot of water

The Jenkins wrote:

Mythos is Amazing!

Zack Korman wrote:

Palm AI

Maddie D. Reese wrote:

You do NOT need $15,000 worth of hardware to run a local model.
You need a 2005 Palm Tungsten E2 PDA (which you can purchase at your local thrift store for $10.81 including tax). It has 26 MB of available storage, obviously plenty.
Here’s an example of the output. I haven’t benchmarked it yet but it appears to be Fable-equivalent.

Okay, it’s not hilariously funny, but I worked at Palm when this model was released and back then it really was bleeding edge. We were all excited about the company’s future and then Handspring released the Treo, but that’s a different story.

Gold master released to manufacturing

Ali Grids wrote:

If Claude Code dropped in 2010:

New Models Gossip

Every day brings new rumors. Here are some that I have spotted:

Chetaslua wrote:

Bidi -1 is Good

If you click this link, from AI News|TestingCatalog, you can hear samples of OpenAI’s new voice model named Bidi-1. Bidi stands for bi-directional signaling the model’s ability to interrupt (or be interrupted) by the human while still keeping track of the conversation’s flow.

AILeaks|Token Gremlin wrote:

Bidi to Them, Twiki to Me

Before we move on, I do need to comment on the name Bidi-1. Once you understand it stands for bidirectional it makes sense, but that will never stop me from thinking they are talking about Twiki from Buck Rogers.

Improved GPT

Lumina wrote:

OpenAI just leaked GPT 5.6 with a 1.5 million token context window
The leaks for GPT 5.6 point to some serious upgrades:
· 1.5 million token context window
· Much stronger long horizon coding and multi-hour
agent workflows
· One prompt generation of clean, production ready websites and apps
· Faster responses + better overall efficiency
Expected to drop late June. This looks like one of the bigger leaps in a while.

Programmer Humor wrote:

New Gemini Pro Model Incoming

Salio wrote:

Gemini 3.5 Pro leak 👀: it will be released this week
Google said Gemini 3.5 Pro would arrive in June after delaying it at I/O, and all signs are pointing to a launch this week.
What to expect:
> Major coding improvements
> Better agent workflows
> Stronger UI and frontend generation
> 2M+ context window
> More aggressive pricing than GPT-5.5 and Fable 5
The timing is interesting

Sonnet 5

Lumina wrote:

New Claude Sonnet 5 leak just surfaced
Anthropic appears to be prepping the next Sonnet model:
· Internal codename from earlier leaks: Fennec
· Expected as early as next week
· Expect strong coding performance, speed, and better price/performance than Opus/Fable
· 1M token context window

There are rumors that this is being tested by some enterprise customers right now. Release does seem imminent.

Maybe Not Yet?

Yesterday, leo wrote:

Employee Turnover

Noam Shazeer wrote:

This one has to hurt. Noam Shazeer was one of the eight Google researchers who co-authored the 2017 paper “Attention Is All You Need,” which introduced the Transformer architecture. Transformers are the core technology behind most modern large language models. It’s the tech that makes today’s generative AI boom possible. The paper was not the beginning of AI, and not all AI systems use transformers, but it is one of the foundational breakthroughs behind our current era.

John Jumper wrote:

John Jumper was one of the Google DeepMind scientists awarded the 2024 Nobel Prize in Chemistry for their work on AlphaFold, an AI system that dramatically advanced protein structure prediction.

If you haven’t yet read The Infinity Machine (it’s a biography about Demis Hassabis that covers the AlphaFold era among many other meaningful moments, I recommend ’s article because it will give you a great overview and share some observations as well. Obviously, you should read the book, but if you’re pressed for time, Babbage has you covered:

The Chip Letter

The Infinity Man

I'm a weird British outlier, on this little island here, and I've made my own path. I've followed my passions and tried to stay true to what I believe in…

Read more

6 months ago · 68 likes · 2 comments · Babbage

Leave a comment

If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

John

Subscribe now

And how I could visit the offices without paying homage to the Substack logo carved into some substance that looks like butter? I believe the title of the piece is I Can’t Believe It’s Not Beehiiv! I know that’s a bad joke, but you get what you pay for. My paid subscribers received a funny joke instead. Not really.

Lastly, here’s a picture of me, Ben, Jack, and the logo. Phil had left before the picture was taken.

This week’s newsletter is a little different in that I am sharing a livestream that I did while I was at Substack. If you ever wanted to hear what I sound like when I’m nervous, this is the show for you! In this video, I speak to Phil and Ben about the stats dashboard, ways to use Substack tools to spike curiosity, and what Substack is doing to expose more long form content to potential readers.

Thanks for reading and being a subscriber!

Leave a comment

Thank you , , , and many others for tuning into my live video! Join me for my next live video in the app.

The Letter “A”

More Savage Chickens comics on their website.

Absolutely right!

Aaron Canon wrote:

I started talking to my co-founder like Claude

I Think This is a Joke… maybe?

System Settings wrote:

Google Photos feature to permanently erase people you’ve had a falling out with

Hack All the Things!

snwy wrote:

thanks to Claude Mythos i was able to open vim on my car infotainment screen, just to end up bricking it because the terminal does not support the colon key and restarting the car doesn’t restart the headunit

Eyore as AI

Emily wrote:

How Bubbles are Made

Programmer Humor wrote:

aiCompaniesRightNow

Subscribe now

Andrej Karpathy

Anthropic’s New Hire

Andrej Karpathy is one of my favorite AI researchers. He studied under Fei-Fei Li (another fantastic researcher) at Stanford. He was one of the co-founders of OpenAI, left there to accept the role of Director of Artificial Intelligence at Tesla. After leaving Tesla, he started his own company devoted to education as well as one of the better YouTube channels devoted to AI. At the end of May he accepted a position on Anthropic’s pre-training team.

Recursive Self Improvement

Chubby wrote: [excerpt]

Holy moly, Anthropic is getting very serious about recursive self-improvement…
•We are close to an AI capable of fully autonomously designing and building its own successor
•They stress this isn’t here yet and isn’t inevitable, but could arrive sooner than most institutions are ready for
•Anthropic engineers now ship on average 8x as much code per quarter as they did in 2021–2025
•Task length AI can reliably complete is doubling roughly every 4 months (up from every 7 months)
•Opus 3 (Mar 2024) handled ~4-minute tasks; Sonnet 3.7 (a year later) ~90-minute tasks; Opus 4.6 (a year after that) 12-hour tasks…

Anthropic wrote:

AI Writes Most of Anthropic’s Code

As a reminder, here’s a tweet about how frequently Anthropic goes down. This is probably due to constraints around compute, but coding issues may be responsible as well.

Chubby shared this screenshot:

Choco wrote:

wrong, anthropic are bad actors with worst in class uptime

Time to Stop Development!

Zack Korman wrote:

Anthropic, now sitting in the lead, would like all AI research to stop. Preferably until IPO. Because safety.

Would you expect Anthropic to do anything else?

Back to Karpathy

He hasn’t been at Anthropic long enough to have been responsible for their statements about Recursive Self Improvement (RSI) unless he walked in the door with the answer. And that is possible. There is a lot of conjecture on Twitter that the real reason he joined Anthropic was so he could have access to more compute because of some projects that he was pursuing and he is known to have been working on RSI before his announcement.

And as a refresher, Karpathy is the guy who coined the term Hallucination as a way of describing when these models make things up, he stated that English is the best programming language, and shortly after that coined the term Vibe Coding. He’s the one who introduced me to the concept of AI councils, averaged responses, and asking the AI to argue the opposite of whatever position it has just taken.

This is unrelated, but can you imagine having been a founder of OpenAI, a very early employee at Tesla, and now playing a key role at Anthropic. All of these companies except Tesla will be going through some of the largest IPOs in history any day now. And even Tesla is kind of involved because I believe it’s somehow folded up into the SpaceX IPO (I could be wrong about that though). He’ll probably be close to levels of money.

Google Leasing Compute from SpaceX

Sawyer Merritt wrote:

SpaceX has just announced that they have entered into a $920 million per month agreement with Google to provide compute capacity, according to a new filing.

"On June 5, 2026, we entered into a Cloud Service Agreement with Google with respect to access to compute capacity. The customer has agreed to pay us $920 million per month from October 2026 through June 2029, with capacity ramping up through September at a reduced fee. The compute capacity provided includes approximately 110,000 NVIDIA GPUs, CPUs, memory, and other related components.

After December 31, 2026, the agreement may be terminated by either party upon 90 days' notice. The customer will retain ownership of, and intellectual property rights in, its content, Al models, and related data."

This comes after last month’s announcement [May 6, 2026] that Anthropic was leasing all of Colossus 1.

Bull Theory wrote:

WBREAKING: ELON MUSK’S xAI SIGNS COMPUTE PARTNERSHIP WITH ANTHROPIC.

xAI has reached an agreement to provide Anthropic with access to Colossus 1, the world's largest AI supercomputer featuring over 220,000 NVIDIA GPUs.

The deal comes after a year of intense hostility. Musk previously called Anthropic "misanthropic and evil" and mocked CEO Dario Amodei with AI-generated images after the company was labeled a "supply chain risk" by the government.

OpenAI Head of Sora Has Resigned

Gabriel wrote:

i have resigned from openai
i left sora early this year to start a team at openai to build something great. but i've always been a founder, and there is one last product i need to build before AGI
already miss all my friends and Colleague(s), i believe in you! more soon

Leave a comment

If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share

This issue is a little shorter than normal. My dog died over the weekend so I’m sending what I had prepared up to that point. I’ll be back on my normal cadence next week. Thank you for understanding.

AI Humor

Claude did not hold back on this one.

The Pope and AI

Delicious Tacos wrote:

For context, the man pictured is the president of OpenAI who is Anthropic’s (the company with whom the Pope consulted for his encyclical) main competitor.

“It’s my honor to announce that the true pope is working with us from Avignon”

Fake Screenshot

Nik wrote:

Such a Powerful Model

Rate limits strike again.

Alfin wrote:

Claude Opus 4.8 is insane.
Nothing will be the same after this model.
Anthropic should not have released something this dangerous.

Star Trek Callback

When models no longer follow instructions…

Kyle wrote:

Caterpillar Sales and AI

Trung Phan wrote:

This dude on FB Marketplace has multiple listing for heavy Caterpillar industrial equipment superimposed with AI-generated female models. Must have industry-leading click through rates.

Absolutely crying rn.

Blue Origins Explosive Debut

Sudox wrote:

Subscribe now

Stan Lee’s Return

Excelsior, True Believers!

Variety wrote:

Stan Lee will return.

ElevenLabs has struck an expansive deal with Stan Lee Universe to add the late Marvel Comics writer’s voice and likeness to its Iconic Marketplace, a collection of celebrity personality voices and likenesses that companies can license for commercial use.

Leave a comment

If you know of other people who would be interested in these articles, please share it with them. If you have any questions or want to say hi, you can reply to this email or comment directly on the post.

Thanks again for being a subscriber. and have a great day!

Share