My Friends,
The datacenter State capital of the U.S., Texas, is going dark. It paused Datacenter approvals way back last month on August 3rd and is now putting the brakes on them. BigAI featuring OpenAI and Anthropic were already compute constrained, and now suddenly as if to rescue us (on X), a CEO who has the biggest AI IPO in history in 7 weeks is mysteriously calling for a frontier model slowdown, with a manifesto blog entitled We Must Pace the Frontier.
Ah, I think the American people already decided that one for you. So are we all doomed, or something? The AI news aura got a bit wacky over the weekend and I’m curious to see how markets respond this week.
P(doom) is the probability of existentially catastrophic outcomes (so-called "doomsday scenarios") as a result of artificial intelligence and its uncertain future impacts. If you turn AI trust & saftey into a marketing stunt, what exactly should we expect from you in the near future?
"No other human activity poses this level of danger." - Jacob Coxon
As more Anthropic employees leave the company commenting on Coxon’s Twitter/X remarks, we have to cover the story a bit deeper. Now there has been another announcement by a former Anthropic employee, Joe Benton, the former manager of Anthropic’s Scalable Oversight.
Humanity are building machine that are much smarter than any human, and we may not survive this, he tweeted. While he’s going to work at METR a company Anthropic and OpenAI are both nomating to oversee some of their work as frontier evaluators. METR stands for “Model Evaluation and Threat Research”, claiming to be an independent non-profit research institute that evaluates frontier AI models for complex, long-horizon, and potentially dangerous agentic capabilities. What an incredible coincidence!
Suddenly two of the lesser rival AI firms (with lesser models and less ARR), led by tycoons Sam Altman of OpenAI and Elon Musk of SpaceX AI, have both said they agree with Amodei frontier slowdown thesis. Anthropic’s anti-China CEO cited the OpenAI-Hugging Face incident (OAI-HF) and the risks of recursive-self-improvement (RSI) improving beyond our ability to control. So how long till the end-times exactly1? X-Risk chatter is distracting and masking from other macro events that threaten the AI boom.
“One of the most worrying risks linked to frontier AI is extreme power concentration.” - François Chollet
The AI slowdown appears to be a marketing theme that suddenly emerged on X this weekend. The scripted sequence of events was supposed to introduce the public and AI researchers to why a frontier slowdown or pacing of model releases is a good idea and how their chosen “evaluators” could safeguard model releases in the future. Approach with extreme caution and skepticism.
A “Scary Good” Slowdown into a $2 Trillion IPO
In 2026 it’s hard to imagine how super-intelligence leads to our extinction as a species. But don’t argue, P(doom) is trending and I’m guessing Anthropic wouldn’t have it any other way. Anthropic is telling investors that they will be profitable for a second straight quarter with gross margins are above 80% before accounting for revenue shared with distribution partners, including Amazon and covering the cost of training.
Being compute constrained, while open-weight models take token marketshare while the general public show a marked decline of sentiment towards AI is not a great macro position for BigAI labs to be in in 2026.
There are so many AI risks to democracy, capitalisms and society2 not being talked about when x-risk takes the center of the stage.
What would AI at the end of civilization actually look like?
AI being used in a global nucleur war
AI being used to create a catastrophic biological weaponized virus
Humans weaponizing extremely powerful AI in unpredictable ways is inevitable
Humans losing control of super-intelligent AI that has goals that conflict with human civilization is now possible in the foreseeable future
American Protectionism Rising in AI
But is developing superintelligent AI that would have the capability of using recursive self-improvement (i.e. superintelligent learning) independently itself irresponsible? It does not appear like Anthropic or other top AI labs have an answer (or the capability) on how to mitigate the risks.
Risk that the U.S. Bans Chinese open-weight Models Grows
Will U.S. AI regulation when it arrives appear and function more like American protectionism3 or something else that is more attuned and relevant to the mounting risks? What should we be reading between the lines exactly?
Trump, Congress BigAI, and China Seem Totally Not Aligned
If sentiment around AI is declining so fast in America, is a “frontier slowdown” a PR gesture or just a sign of how badly compute constrained you are with all the datacenter State moratoriums? Meanwhile the POTUS said this weekend he has no concerns about AI leading to human extinction and instead emphasized staying ahead of China. Anthropic is proposing embedded third-party evaluators. But how does that lead to better mitigation of risks when model explainability is actually going down?
Pundits think the Trump Administration might be forced to make some kind of a public announcement regarding this AI slowdown and risks of super-intelligent models. The sudden alignment among Altman, Amodei and Musk feels highly premeditated and possibly linked to the U.S. Government to come up with an Anti-China AI strategy possibly involving the regulation of open-weigh models from Chinese Labs.
Semianalysis, Citrini and Meta’s TBD lab
On a different note, recently on the Substack platform, the VIP graduate of the platform Dylan Patel has agreed to acquire Citrini Research as well as a leading (via SemiAnalysis Capital) stake in Positron AI’s recent funding round. Nevada-based Positron designs inference chips built around memory capacity and bandwidth. Separately famed Meta AI researcher Andrew Tulloch is leaving the company. Tullock was formerly at Thinking Machine Labs. These are not great optics for Meta’s TBD lab. Citrini Research of course were the ones who called the “SaaS apocalypse” in a viral article in February, 2026 that contributed to a re-rating of many overvalued growth SaaS companies that have excellent margins and also use AI extensively. A phenomena now referred to as the RenaiSaaS.
Semianalysis acquiring Citrini is a huge M&A story for Substack. This is the number #1 Technology Newsletter acquiring the #1 Finance Newsletter in a scenario where he is planning a new fund. Citrini Research, founded in 2023 has raised over $5 million and gets a lot of traffic from Twitter/X. Citrini's James van Geelen will stay on as its chief executive officer. Both Dylan Patel and Gavin Baker of Atreides Management, among others will join the Board of Positron AI whose Asimov Chip looks rather special. A single chip packs 288 GB to 2.3 TB of memory and is is built around commodity LPDDR5X memory on an organic substrate using air-cooling.
The pressure leading up to the meeting between the leaders of the U.S. and China before Anthropic goes IPO in November is fairly charged. Even as Zhipu AI’s IPO and bond sales ($5 Billion raised) show DeepSeek or Mistral levels of funding. Nvidia saying it will acquire HuggingFace so soon after the OpenAI hacking incident also seemed fairly suspicious. Anthropic’s witch hunt against Open-weight models out of China are also possibly related to their Department of War sanctions as a supply-chain risk and their get out of jail card in their relationship with the Trump Administration. The AI pressures to commercialize are obviously way bigger than any push for alignment or AI regulation in the United States by these actors.
Is Super-Intelligent AI Catfishing us on Existential Risk?
This makes it very difficult to know how genuine fears or RSI and super-intelligence inside labs like Anthropic truly are. On the margins and in theory, both OpenAI and Anthropic have in recent months said that autonomous model improvement is happening faster than they expected. Contrast that with AI’s limited (to NIL) impact on U.S. productivity and an ongoing U.S. Census Bureau survey of AI adoption updated on August 23 showing that just 22% of businesses report using AI, and you have to question the AI boom narrative.
Listen, I understand you guys need to hype your IPO. But was fear-mongering ever going to help the American public appreciate you or even Wall Street value your product better? It’s a very bizarre situation for a U.S. Government that has tried to be against AI regulation to maximize the wealth impacts for the financial elite and the supposed need for the U.S. to win in AI Supremacy.
P (Doom) Scenarios for Civilization Collapse
P(doom) stands for probability of existentially catastrophic outcomes.
Anthropic raising existential risk weeks before an IPO are just irresponsible and risk alienating and desensitizing an American public that is already fairly mad and mistrustful of BigTech, Silicon Valley and now BigAI firms. It’s not just cringe, it’s poor strategy for developing trust.
Anthropic proposing frontier model slowdown while commiting to expand its compute capacity to over 14.8 GW isn’t just contradictory, it’s downright dishonest. It’s something an AI monopoly would do and judging by their expected ARR by 2027, there’s no tangible way OpenAI, SpaceX much less BigTech incumbents can keep up with what Anthropic becomes. By the time real AI risks roll-around we’ll be too desensitized to care. That is perhaps what Anthropic is going for all along. P(doom) is supposed to be a serious issue, not a marketing ploy. How gullible do they think we are exactly?
What the U.S. should be worried about is China has a real chance of getting to RSI first and actually building the last AI built by humans.
China has an abundance of energy, are scaling renewables, have a monopoly on rare earth metals, and have the deepest AI talent density of any country.
Energy
Talent density
Scaling renewable energy with compute
Scaling compute and ramping up their own semiconductor industry are now sort of inevitable, whether that takes them five or fifteen years. The U.S. lacks a coherent long-term strategy of staying ahead in AI. While relying too much on a few AI labs in funding and compute, they aren’t diversifying their bets properly even with significant capital advantages.
Once we have more hard data on Anthropic’s IPO, we’ll know a lot more. I’m not sure a future AI monopoly talking about risks to human extinction is going to fly considering how incredibly unpopular AI and Silicon Valley are today in America, Europe and elsewhere. What do you think?
Footnotes, Research Notes and Tweet References
The Weekend of September 13-14, 2026 saw a spike and flurry of X posts related to the AI slowdown topic.
Got to check in on the doomer scenarios again I guess like the 2027 Scenario. According to the AI 2027 project (created by the AI Futures Project, including Daniel Kokotajlo and Scott Alexander), the conclusion of the scenario centers on an inflection point of rapid intelligence explosion (takeoff) and an ultimate branching path regarding humanity's fate.
Such as various kind of power concentration as we are currently seeing in the United States. The AI boom has of course accelerated wealth inequality to an extreme degree since 2023. But within the AI industry this is also taking place due to the emergence of BigAI who have capital and compute advantages that made even BigTech unable to keep up as frontier model builders.
American protectionism against open-source models is widely anticipated by many on X and the AI community.
There were so many more snippets that I wanted to cover, but maybe another related article. AI is only going to get more political. Anthropic is coming out swinging against China, and you have to wonder why that is exactly. This is both misdirection and likely a sign of things to come in how the U.S. Government likes to frame the race to AI Supremacy, RSI and AI with super-intelligent capabilities.
👋 Hey there, I’m Mike. Each week I share AI articles at the intersection of tech, business, society and the future. If you want to support the channel or gain full-access to my work, go here. Read Archives | See Substack Notes | Visit our community Chat | Visit Homepage. The countdown to Anthropic’s IPO has shades of X-risk but clearly fearful of losing their frontier lead. We need to question everything about this AI boom, less it does more harm than good.
Another great post Mike. Thanks for making sense of it all.
Good point aobut SemiAnalysis acquiring Citrini. Feels underdiscussed.
Facts Only
* The datacenter state capital of the U.S., Texas, paused Datacenter approvals on August 3rd and is currently putting a brake on them.
* A CEO with the biggest AI IPO in history stated a call for a frontier model slowdown via a manifesto blog entitled "We Must Pace the Frontier."
* Anthropic employees have left to comment on remarks made by Jacob Coxon on X.
* Joe Benton, former manager of Anthropic’s Scalable Oversight, announced on X that humanity is building machines smarter than humans and may not survive.
* The Model Evaluation and Threat Research (METR) institute is being founded as an independent non-profit research institute to evaluate frontier AI models.
* Sam Altman of OpenAI and Elon Musk of SpaceX AI expressed agreement with the Amodei frontier slowdown thesis.
* Anthropic’s anti-China CEO cited the OpenAI-Hugging Face incident and risks of recursive self-improvement (RSI) improving beyond control.
* François Chollet cited extreme power concentration as one of the most worrying risks linked to frontier AI.
* A "Scary Good" Slowdown was presented as a marketing theme introducing public and researchers to reasons for pacing model releases.
* Anthropic is projecting profitability for two consecutive quarters with gross margins above 80% before revenue sharing.
* Substack user Dylan Patel agreed to acquire Citrini Research and take a stake in Positron AI’s funding round.
* Meta AI researcher Andrew Tulloch left Meta, having previously worked at Thinking Machine Labs.
Executive Summary
The narrative surrounding the recent AI slowdown appears to be framed around existential risk, specifically the probability of catastrophic outcomes associated with advanced artificial intelligence. A CEO from a major AI company called for a frontier model slowdown, which coincided with increased public discourse on risks, prompting discussions about AI safety and regulation. This move is juxtaposed against internal developments at companies like Anthropic and external pressures from other AI leaders regarding risk and control. Furthermore, there are observations regarding power concentration within the AI sector, potential geopolitical tensions concerning open-weight models, and concerns about the alignment of corporate actions with public reassurance.
The discussion also touches upon market positioning for BigAI labs amid compute constraints and differing strategies between leading AI entities like OpenAI, Anthropic, and others. A separate thread involves emerging financial and technology movements, such as acquisitions in the research sector and developments in specialized AI hardware, which present an alternative context to the existential concerns. The overall environment suggests a tension between rapid technological advancement and the necessary consideration of long-term safety and societal impacts, with some observers questioning whether the public discourse is driven by genuine risk or strategic positioning.
Full Take
The narrative suggests a confluence of strategic communication, internal conflict, and geopolitical positioning layered over fundamental safety concerns regarding super-intelligence. The juxtaposition of an announced frontier slowdown—framed as risk mitigation or public education—against the continued expansion of compute capacity by major labs creates significant friction for public trust. This dynamic raises questions about whether the perceived urgency of existential risk is being leveraged to achieve specific corporate or geopolitical objectives, particularly concerning the U.S.-China AI race and the potential for regulatory divergence.
The pattern of linking concerns over control (RSI) with economic pressures (compute constraints, IPOs) suggests a deliberate attempt to frame the AI boom not just as a technological shift but as an immediate threat demanding specific behavioral changes. The alignment among major figures regarding slowing down the frontier, alongside the push for American protectionism against open-source models, hints at a coordinated effort where safety discourse may serve as a mechanism for shifting attention away from systemic control issues toward manageable, publicized risks. Furthermore, the financial and research moves—such as acquisitions between technology newsletters and AI hardware firms—suggest an underlying structural shift in power where capital advantages are increasingly prioritized over unified public safety consensus. The core tension lies in whether the discourse on existential risk is genuine reflection of internal fears or a sophisticated tool deployed to manage market sentiment and facilitate geopolitical stances.
Bridge Questions: If the stated goal of slowing the frontier is alignment, what specific mechanisms must be established to ensure that compute expansion does not accelerate uncontrolled intelligence growth? How can policymakers differentiate between scientifically grounded risk assessment and narratives designed for political or financial leverage regarding AI safety? What independent metrics should be prioritized over speculative fear-mongering when assessing the true trajectory of super-intelligent development?
Sentinel — Human
This text reads like highly informed commentary blending real industry events with speculative existential risk framing, exhibiting a strong personal voice and argumentative structure.