- WKND AI
- Posts
- Trump Shuts Down Mythos
Trump Shuts Down Mythos
+24 Year Old Manages a $20B AI Hedge Fund
In partnership with
AI agents now read your docs almost as much as humans do
5% of traffic to your docs is now AI agents, not humans. If your documentation isn't structured for machine readability, your product is invisible to Claude, Cursor, and every other coding agent your buyers use daily. Mintlify is built for both audiences.
⬆️Support this newsletter by clicking on the link above ⬆️
Hello WKND AI Warriors!
FP&A leaders are getting tired of AI “time savings” and want one thing instead: ROI.
Also, Anthropic spent months warning the government about dangerous AI models, only to complain when regulators restricted access to its own flagship release.
Plus, critics say Anthropic got exactly the AI oversight it spent months lobbying for, then immediately pushed back when it affected its own models.
Oh yeah, and one former OpenAI employee turned a $225M AI hedge fund into a $20B giant in just two years.
Today’s newsletter includes:
💼 AI AROUND THE OFFICE
📰 AI NEWS RECAP
🤿 AI DEEP DIVE
💼 AI AROUND THE OFFICE
CFOs Want ROI, Not Just Time Savings
Background
I met with a CFO whose mindset mirrored the Fortune 100 story: hundreds of thousands spent on AI, zero visible ROI. They weren’t impressed by “we’ll shave 20% off your month-end close”; they wanted concrete impact on cash conversion cycle and return on capital.
Pain Point
AI initiatives are stuck in “productivity theater” (faster decks, faster commentary) while leadership wants improvements in working capital, margins, and capital efficiency. The gap between pilot-level wins and P&L-level impact is still wide.
Root Cause
Most AI use cases are local optimizations, not tied to revenue, margin, or CCC.
FP&A isn’t always in the room when AI priorities are set, so projects don’t map to financial levers.
Data needed to connect AI use cases to actual cash or capital metrics is fragmented.
Resolution
Reframe FP&A AI from “time savings” to “unit economics and cash.” Only pursue AI use cases where you can draw a clean line to DSO, inventory days, DPO, gross margin, or operating expenses.
Next Steps (What You Can Do Today)
Write down your top 3 financial levers (e.g., reduce DSO by 5 days, cut non-headcount opex by X%).
For each AI idea, force the question: “How does this move one of those levers?” If you can’t answer, deprioritize it.
Run one simple baseline: pick a process (e.g., collections outreach) and measure current metrics before any AI changes.
Design one AI experiment with a clear, measurable target (e.g., AI-generated dunning emails to reduce DSO by 2 days in one segment).
📰 AI NEWS RECAP
Anthropic continues to butt heads with the Trump administration.
For months, Anthropic has been warning everyone that their Mythos-class models possess dangerous, automated hacking capabilities, effectively advocating for strict government regulation of frontier AI.
To offset the risk, they initially limited access to a closed cohort of trusted tech and cybersecurity companies under an initiative called Project Glasswing to test infrastructure for vulnerabilities.
Then, they deployed a commercially restricted version called Fable 5 to enterprise customers and consumers, believing they had aligned with government safety institutes.
But then came the twist.
One of those testing companies accidentally tipped off the government that they were able to jailbreak the model.
The kicker?
It was Amazon who has 8 billion reasons (the cash they’ve invested in Anthropic) to keep quiet.
Amazon CEO Andy Jassy casually mentioned the exploit during an unrelated White House phone call.The government panicked.
Citing national security, the Commerce Department enacted aggressive export controls restricting Mythos and Fable access to all foreign users.
Because these rules applied even to non-US citizens working inside Anthropic, the company was forced to abruptly disable Fable 5 and Mythos 5 commercial endpoints globally just to remain compliant.
Oddly enough, Anthropic’s CEO, Dario Amodei, then immediately pushed back, arguing that a narrow technical jailbreak shouldn't justify recalling a commercial model.
Pundits are having a field day calling out the blatant hypocrisy commited by Dario:
Riling everyone up about how dangerous the tech was.
Publishing essays demanding the government have the power to block unsafe models.
Complaining the second the government does exactly what he called for.
You couldn’t make this up if you tried!
🤿 AI DEEP DIVE
Every week people ask me how to make money in AI.
It’s tough, but one easy way?
Follow the investment thesis of Leopold.
Don’t know who Leopold is?
He’s a 24 year old former OpenAI employee who started an AI hedge fund that has grown from $225M to $20B in 2 years.
Oh yeah…and I featured this video about him TWO YEARS ago.
Just another reason why you should share this newsletters with someone you know!
