The video argues AI remains an earnings machine even as the political and regulatory backlash accelerates: SpaceX keeps stacking compute revenue, frontier labs keep pushing larger models, and the market keeps paying for infra, chips, and distribution. At the same time, Anthropic’s own modeling is used to frame a brutal labor outcome — richer GDP, lower wages, higher unemployment, and a collapse in knowledge-worker share. The last stretch shifts to politics, media, and safety: X as a dominant news source, ABC, AOC, the UK’s chip/superintelligence bill, and warnings from Jeffrey Hinton and Elon Musk about control and existential risk. The through-line is that AI may create huge profits first, but the social and regulatory fight is only getting sharper.
SpaceX added an unnamed deal believed to be with MSFT worth $1.1B/month, or $13.3B/year, lifting its high-margin compute commitments to $41.1B.
SpaceX is already 41% of Elon Musk’s $100B ARR target for year-end.
Grok 4.7 is counting down to a teased 2.1T-parameter model, with Musk claiming it will exceed every current model, not just Grok models.
Astra launched GPT-6 in the last 7 days, called it the most intelligent and aligned model, and claimed 99.9% on ARC-AGI.
Astra’s founders said GPT-6 is AGI and that AI knowledge work can exceed human capability.
DeepSeek’s new model is said to match Astra on some benches like design while costing 1.4% of the price, and Oracle reported 98.7% GPU utilization.
Google is portrayed as backing away from frontier models toward infra, while SpaceX is doing both infra and frontier models and frontier labs are building chips to push further into infra.
Meta is focusing on personal agents for calendar management, travel booking, email, form-filling, and possibly shopping, distributed through mobile and WhatsApp.
Kevin cloned bots of Sam Altman, Elon Musk, Dario, and Zuck as Grok bots, locked them in a chat cage, and got them to debate AI risk until they signed a joint statement rejecting any one government controlling AI’s benefits.
Anthropic said Iran used Claude to target US Navy-related positions, work, and files, and also to run influence campaigns.
Anthropic’s economics study modeled extreme-path GDP up 32.4% by 2030, with the substantial case +8.3% and the modest case +1.6%.
In Anthropic’s extreme path, knowledge-worker employment share falls from 60% to 45%, headline unemployment rises from about 4% to 12%, knowledge-worker unemployment hits 18%, knowledge-worker wages fall 11%, and labor share drops from 60% to 45%.
US tech is said to be posting record profit margins, with S&P 500 profit margins at 24%, described as the highest in history, and the market keeps paying for AI picks-and-shovels because demand for intelligence is framed as effectively infinite.
X is claimed to be the #1 source of news worldwide across 167 countries after Musk bought it, while ABC News was said to have been ratioed and community-noted for a Tesla-scaffolding crash claim that was later deleted after challenge.
On September 8, Jacob said he resigned from Anthropic after 3 years doing pre-training research at OpenAI and Anthropic; the notes also say he had only been at Anthropic 6 weeks, Musk called it a setup or sci-op, and Jensen Huang said it was crazy.
Jeffrey Hinton appeared on BBC Newsnight the same week Astra was pushed and said there is a 10% chance AI kills all humans.
On September 9, a UK Private Members Bill was introduced in the House of Commons proposing security-bans rules for artificial super intelligence and tighter chip-supply oversight.
AOC was accused of implying white people were hanging black people after a claim of 10 reported hangings of people of color in the US this year; Musk used Grok to list cases by the murderer’s race and said there were only nine suicides and one murder by a black person.
The closing frame is that politicians of a certain ilk will try to destroy AI, even as AI is treated as the lever for exposure and investment gains.