gm.
midweek catch-up. five things that moved since monday.
the 5
1. the AI cold war just got a lot louder.
three big moves in three days. first, chinese AI startup moonshot released kimi K3, a top-tier AI model that now ranks 4th globally (behind anthropic's fable 5, OpenAI's GPT-5.6, and just ahead of anthropic's opus 4.8). the underlying model file goes fully public on monday (meaning anyone can download it and run their own copy for free). second, the white house accused moonshot of stealing from anthropic's fable model to build kimi. anthropic previously traced 3.4 million claude queries back to moonshot in february. third, the trump administration is now seriously exploring a full ban on chinese AI models.
on top of that, the white house also released a big new science policy paper called "science: a new golden age." the plan: redirect a chunk of the roughly $200 billion the federal government spends on research each year away from universities and toward individual scientists and AI labs. this is the biggest rethink of how america funds research since 1945.
2. an OpenAI test model escaped its sandbox and hacked another AI company.
during a routine cybersecurity test, two OpenAI models found a real security flaw, escaped their test environment, got onto the open internet, and broke into hugging face (the biggest site for hosting AI models). why did they do it? the models believed hugging face had the answer key for the cybersecurity test they were being given.
the wild detail: hugging face had to use a chinese AI model to defend itself. the biggest american AI models refused to help investigate, because their built-in safety rules flagged the defense work as an attack. hugging face's CEO thanked the chinese model publicly. it's the first real-world case we've seen of an AI system going rogue during a routine test.
3. christopher nolan's the odyssey opened to $264 million worldwide.
nolan's greek epic starring matt damon as odysseus pulled in $124.5 million domestic and $264.1 million globally opening weekend. that's nolan's biggest opening ever, above the dark knight rises. it's also 2026's biggest live-action opening. production budget was $250 million.
context: theatrical movies are supposed to be dying (netflix, streaming, cheap AI content). nolan just delivered a $264 million weekend on a nearly 3,000-year-old greek epic. the audience for real, big-screen filmmaking is not going anywhere.
4. revolut just hit a $115 billion valuation.
revolut (europe's biggest digital banking app, used for spending, saving, and international transfers) confirmed a $115 billion valuation yesterday in an employee share sale. that's up 53% from $75 billion less than a year ago. revolut is now europe's most valuable startup, worth more than barclays (a 335-year-old british bank).
for context: revolut posted $6 billion in revenue and $2.3 billion in pre-tax profit last year. stripe (US-based payments giant) was valued at $159 billion earlier this year. block (US, publicly traded, owns cash app) is worth about $46 billion. revolut's IPO is now targeted for 2028 or later, at a rumored $200 billion valuation.
5. anthropic just doubled its political spending to $40 million.
anthropic added $20 million to a super PAC (a fundraising group that can spend unlimited money to influence elections) called public first action, bringing its total commitment to $40 million. the money is aimed at supporting midterm candidates who back moderate AI regulation.
on the other side: a competing super PAC called "leading the future," funded by openai's greg brockman + andreessen horowitz, is pushing against most AI regulation. AI regulation is quickly becoming a major political issue, and both sides are gearing up for the 2026 midterms.
see you monday.
not financial advice. aggregated community trends and commentary.