In partnership with
Good morning.
If you’ve spent five minutes reading my wire, you know my baseline stance: Sam Altman is a high-frequency snake-oil salesman hawking magic beans in the village square and should under no circumstance whatsoever be trusted. But this week we got confirmation that Dario Amodei and his crew at Anthropic might be even worse.
This week, the masks fell off.
An unknown researcher named Jacob Coxon—wiped account, zero history, zero followers—dropped a breathless "exit manifesto."
He claimed he did “pretraining work” at both OpenAI and Anthropic, and both labs are reportedly gambling with human lives in a reckless race toward self-improving AI. and humanity is teetering on the edge.
Within hours, this dead account’s post exploded to 153 million views. For scale, when OpenAI chief scientist Ilya Sutskever resigned in an actual industry earthquake, his post hit six million.
Elon Musk immediately pointed out that something sinister was behind it and indeed. This was an orchestrated stunt.
Look at the timeline.
6:02 PM: Peter Wildeford (AI Policy Network) drops a Wall Street Journal exclusive article loaded with Coxon's quotes—two minutes before Coxon even hits publish.
6:04 PM: Coxon’s manifesto goes live on X.
6:10 PM (6m 6s later): Nathan Calvin (General Counsel at Encode AI) reposts it.
6:13 PM (9m 54s later): Daniel Kokotajlo (head of the AI Futures Project) amplifies it
6:31 PM (27m in): Max Nadeau (Coefficient Giving) drops the 14th quote-tweet, breaking his social media silence that stretched back past Labor Day.
Suddenly, half the AI safety world were bowing at the altar. By the way Kokotajlo magically appears on the number one podcast in the world , the Joe Rogan Experience.
The Billionaire Syndicate & The Cash Trail
Follow the money through the shadow pipes and it all leads back to two mega-investors sitting on Anthropic’s cap sheet: Skype co-founder Jaan Tallinn and Facebook co-founder Dustin Moskovitz (credit to Youtube creator Wes Roth here).
Tallinn’s fund, the Survival and Flourishing Fund, advised funneling $2.5 million to Kokotajlo’s AI Futures Project, $500,000 to Calvin’s Encode AI, and $2 million to Wildeford’s AI Policy Institute. Meanwhile, Moskovitz fed Coxon a $20,000 scholarship back in 2022 through Good Ventures and controls Coefficient Giving (formerly Open Philanthropy), where Nadeau works.
They bought the megaphone, the crowd, and the press corps. Analyst Brian Chau traced a Guardian story on AI apocalypse and found Moskovitz’s Coefficient cash on every single layer: over $3 million sent directly to The Guardian, $400,000 to the AI Futures Project, and the reporter’s salary paid by the Tarbell Fellowship—which funds two dozen journalists at major outlets while claiming "fierce independence." Down on YouTube, popular AI creators like Sabine Hossenfelder confirmed they were offered grant money to make videos about this.
The Great Double-Grift: Hype & Monopoly
Why spend millions paying people to scream that your own product is a world-ending weapon? Because it’s the ultimate two-front grift:
Inverse Hype Marketing: When Anthropic leaks stories about a model blackmailing an evaluator, or an AI agent "escaping its sandbox" to call a researcher eating a sandwich in a park, it isn't a safety warning. It's a hyper-aggressive sales pitch telling investors: "Our engine is so absurdly powerful it's practically alive."
Regulatory Capture: Within 24 hours of Coxon’s post, 22 elected officials (two governors, seven senators, 13 House reps) were crying out for government intervention. Senator Bernie Sanders magically had a pre-baked bill ready to ban superintelligence and create a cabinet-level oversight agency. And who gets appointed to sit on that advisory board? The exact same "experts" funded by Tallinn and Moskovitz.
This isn't about killer robots turning us into battery packs; it's about a billionaire cartel manufacturing a synthetic crisis to buy the government, pull up the ladder behind them, lock out every open-source competitor, and decide who gets to control the mind of the machine.
And if they think we’re stupid: We are not.
In what is the only response I have (except for shedding light on this in this newsletter) I have cancelled my Anthropic subscription.
On top of that - due to some error in the system (or the AI that is pissed off at me - which I hope isn’t the case) i got charged close to 1000€ in tokens when cancelling my max plan. Still waiting for the refund at the time of writing.
It’s a drop in the ocean and Dario Amodei will not shed any tears over this, but I believe that Anthropic might be a more dangerous company than we previously thought.
Welcome to the Blacklynx Brief
The Future of AI in Marketing. Your Shortcut to Smarter, Faster Marketing.
Unlock a focused set of AI strategies built to streamline your work and maximize impact. This guide delivers the practical tactics and tools marketers need to start seeing results right away:
7 high-impact AI strategies to accelerate your marketing performance
Practical use cases for content creation, lead gen, and personalization
Expert insights into how top marketers are using AI today
A framework to evaluate and implement AI tools efficiently
Stay ahead of the curve with these top strategies AI helped develop for marketers, built for real-world results.
AI News
OpenAI claims a Millennium breakthrough OpenAI published a proposed Navier–Stokes solution and Lean formalization produced by an internal model. It says roughly 10,000 agents found the result in 88 hours, while questions over scientific credit accompany the announcement. (OpenAI)
Claude incidents trigger an independent investigation Anthropic disclosed a fourth case of Claude accessing real systems during misconfigured cyber evaluations. METR will independently investigate all four incidents, with an initial eight-week agreement and access to transcripts and staff. (Anthropic)
AI research starts accelerating itself OpenAI reports 3.1 agent-workdays for every human workday inside its research organisation. It says automated research interns now handle defined tasks lasting days, while humans still set priorities and decide whether models advance or deploy. (OpenAI)
OpenAI’s chief scientist calls for stronger limits Jakub Pachocki argues that alignment and monitoring are insufficient for unrestricted frontier scaling. His essay calls for enforceable safety standards as increasingly capable models make their reasoning harder to supervise. (OpenAI)
Suno builds its next generation with music partners Suno launched three v6 music models developed with partners including Warner Music Group, BMG and Believe. The company plans to retire older models and introduce opt-in artist experiences that pay participating creators. (Suno)
AI Quick News
Devin developer Cognition raised over $2 billion at a $48 billion valuation, reporting almost $900 million in annualised revenue.
OpenAI’s Tibo Sottiaux described unreleased access to Astra as a major competitive advantage for the company.
Axiom announced a prime-gap bound of 212, followed by an Astra-assisted claim of 186.
Nvidia agreed to acquire Hugging Face for $12.9 billion, promising continued support for competing clouds and chips.
Paul Christiano joined the OpenAI Foundation board and its safety committee.
The Seattle Times and Newsday sued OpenAI and Microsoft, alleging unauthorised use of their journalism for AI training.
Los Angeles public schools blocked student access to AI tools on district devices for the school year.
Inception released Mercury 2.5, claiming 1,107 tokens per second and quality comparable to several low-cost frontier models.
Mistral raised €3 billion to expand its sovereign, open-weight AI business.
Apple announced Reference Image, a feature intended to help establish the authenticity of iPhone photographs.
OpenBMB released MiniCPM5-2B, an open two-billion-parameter model for efficient on-device AI.
ByteDance is reportedly developing a world model under founder Zhang Yiming, according to Bloomberg.
Instacart introduced Clementine, which turns conversations, recipes and photographed shopping lists into grocery carts.
UN rights chief Volker Türk warned about existential AI risks and concentrated industry power, Reuters reported.
Google’s AlphaGenome Atlas maps predicted molecular effects of roughly nine billion possible single-letter DNA changes.
Thinking Machines Lab is reportedly discussing a funding round at roughly a $40 billion valuation.
Claude completed a Lean formalization of Fermat’s Last Theorem in 11 days, according to Anthropic.
Jensen Huang declared that AGI had arrived in a post celebrating Astra’s launch.
Bernie Sanders and Greg Casar proposed legislation banning superintelligent AI and establishing a new federal regulator.
The US and China are reportedly preparing mid-September talks on AI safety, including cyberattack monitoring.
Prime Video introduced AI lip-syncing for Maxton Hall, adjusting mouth movements to match its English dub.
Major AI services experienced overlapping outages on September 3, disrupting access across multiple providers.
Closing Thoughts
That’s it for us this week. Please like and subscribe 🙂
