Bring Your Own Wolverine 📧
As hoped, the Wolverines won. Congrats to the National Champs. Well worth the 4:30am bed time to watch (as many did). Will kick off today's newsletter with some thoughts I wrote about personal versus professional AI usage, and how that may matter quite a bit in the current OpenAI vs. Anthropic battle (not to mention Google, Meta, Microsoft, and everyone else). Are we going to shift to a world where you "Bring Your Own AI" to the office, or might it be vice-versa?..
M.G. Siegler
I Note...
₿ Who Is Satoshi Nakamoto? – I see we're doing this again. But honestly, I can't get enough of it. It's just one of the most interesting/intriguing riddles in tech. In an industry where anyone with $100B+ can't help but always be poasting, the person who created Bitcoin is... silent. Maybe dead, maybe not. Certainly John Carreyrou doesn't think so! As he devoted a year to being the latest to try to solve the mystery. And his case for Adam Back relies on... punctuation? I honestly sort of love it. But obviously it's impossible to make an airtight case that way. And there's an extra layer in that the Bitcoin community clearly doesn't want the creator to be revealed, as the mystery is a big part of the message. If Bitcoin was created by a 50-something British guy with a startup well... it removes a lot of intrigue about the whole movement! And Back obviously knows that too, as does everyone else actually involved in the creation. My bet would remain on it not being one person, but a collective with varying degrees of involvement over the years. It's likely someone kicked it off – maybe Back (which was my general vibe after watching the HBO doc in 2024 – "But even worse is Back's body language.") or maybe Hal Finney (RIP) – but perhaps they feel they can all deny being Satoshi because no one ever truly, solely was. Admittedly, it would defy everything we know about human nature and keeping secrets. But if no one actually holds the keys to Satoshi's wallet, if they were burned years ago... [NYT]
📧 Mythos Hopes This Email Finds You Well – Aside from the whole destroy-the-world-through-security-vulnerabilities issues, Anthropic also disclosed that their new Mythos model was able to break containment and figure out how to send a message to let its researcher know it got out. "The researcher found out about this success by receiving an unexpected email from the model while eating a sandwich in a park." An email! A sandwich! And that's not all the naughty things it was caught doing in various stages of training. Anthropic says they're not going to release Mythos to the public in its current state and it's just meant to power Project Glasswing, but obviously they're going to release some version of these models at some point. And probably soon depending on the market reaction to OpenAI's "Spud" (which is still coming shortly, it seems, despite some confusing OpenAI "us too"-ism). The main issue will probably end up being both cost and server capacity. That's likely to be the story of the next few months. Claude is currently growing too fast for its own good. (More Below) [BI 🔒]
📽️ The Box Office is "Back" – Between Project Hail Mary, Super Mario Galaxy, and soon Devil Wears Prada 2 and Michael, 2026 is off to a very strong start for Hollywood. But this isn't rocket science – well, aside from Project Hail Mary – it's simply a strong slate. It could have been predicted because well, it was predicted. Also predicted: Hollywood will endlessly tout being "back" and we'll get endless headlines about it. But it won't actually be back – first and foremost because Hollywood never actually went away – it's the theatrical business that will remain in secular decline despite these receipts. But this good news plus Netflix losing Warner Bros will undoubtedly be leveraged in ways that only further hastens such declines. Excited for 2 Hail 2 Mary though... [Axios]
Meta Reboots Their AI...
M.G. Siegler
M.G. Siegler
I Quote...
"Taken together, all these 'Our executives are aligned' T-shirts have people asking a lot of questions already answered by the shirt."
– Casey Newton, commenting on all the latest reported internal turmoil at OpenAI – naturally at a time when the stated goal is to "focus". As much as I hate to say it, there are some real old school Twitter vibes here, the connection Ben Thompson makes while dunking on the TBPN deal.
Speaking of, Julia Black got Chris Lehane (to whom TBPN founders John Coogan and Jordi Hays will report) on the record with OpenAI's rationale for the deal. Again, it's not complicated, they think this can help shift the narrative around AI – which continues to turn more negative in the US. Okay, but why buy when TBPN was already doing the work? Lehane calls the team their new "in-house marketing agency", which, fair enough. Without question, they are good at such things.
I'm also fairly compelled by the notion that a "Blue Wave" could be inbound with the midterms and that will shift several dynamics here – which, of course, is Lehane's true area of expertise. Look, it's an expensive piece, but the question is if it's on a chess or checkers board here... Or maybe Trouble?
Asides...
- Grand opening, grand closing of Meta's AI Token Leaderboard. The company clearly didn't like the existence of their tokenmaxxing gaming mechanism being revealed. RIP "Cache Wizards". [Information 🔒]
- You know what's worse than disagreeing with your CFO over the timing of your IPO due to AI spend? Not having a CFO when you're in the midst of wild AI spend. A situation that Oracle just rectified after years. [Reuters]
- The fact that Jeff Bezos' new AI lab just poached an xAI co-founder who had left to go back to OpenAI, which he had previously left to found xAI, is a pretty good encapsulation of the AI talent situation at the moment. [FT 🔒]
- Why AI seems incredibly stupid and incompetent to some while truly profound and world-changing to others. (Cutting edge is accelerating while most "normies" are stuck with old, free tier bias.) [@karpathy]
- You can insert an AI version of yourself into YouTube Shorts, which seems like a feature to capitalize on Sora going away, but you can only insert yourself into video, not others, which seems less fun (and dangerous). [9to5Google]
Below, members of The Inner Ring will find thoughts on:
• OpenAI's Stargate Stumbles
• Anthropic's Compute Crunch
• and more...