How's your Black October going? We're fine (promise), frantically patching everything we can get our hands on while every risk indicator in the world lights up. Also, the age of personal super-negotiators is here! youtube.com/watch?v=5_yu50ES…

“we are 100% relationship people” Clearly 💀

God bless confident idiots with hot takes! (The demo video literally opens to the Stripe UI for approval)

This is a good thread. By the time you understand every statement in here, you’re better equipped than most to reason about AI as an industry.

Social media provides such a valuable, engaging forum for meaningful conversation and learning.

Dario is trending right now for saying that maybe tokens should be charged based on the output value, since some tokens might be "worth more" depending on what they're used for. And he's right, of course, in that some uses of a commodity (like charging a phone to watch cat videos) is ultimately worth less than others (like powering up a CNC machine to make expensive parts). Same electrons, different market value. But that's just an argument for Anthropic being a vertically-integrated technology company: They have the models, they're acquiring the compute, they can charge based on output value for absolutely anything they want. They could defeat Office, Google Workspace, Salesforce, NetSuite, and any other software product by simply building and hosting their own - AI makes that possible, right? They could defeat manufacturing by designing better factories with AI, programming the firmware with AI, building the backoffices and distribution with AI, going to market with AI. They could defeat pharmaceutical companies by researching better drugs using AI, building their own supply chains, even doing their own last-mile with self-driving/drone delivery, if they want. They could defeat pretty much all of banking, investing and financial services, 90%+ of which is digital work by now. They could build better banking apps, better banks, better insurance models, better everything. They could *definitely* defeat social media, marketing and advertising - the software for hosting platforms like Facebook, Instagram, X, QQ and the rest isn't technically sophisticated, you just need tons of infrastructure, and they have the money for it. So why don't they? Why is Anthropic's S-1 94 pages of existential risk, questionable numbers, and a requirement to raise more capital even though they have large amounts of cash on hand? Why aren't they just rebuilding entire industries? They could ship a world-class operating system, world-class productivity software, world-class business systems for (as their "country of geniuses" argument would suggest) a fraction of what it costs humans to do the same. Because if it makes sense to charge for compute based on the output value, you're going to have a hard time finding customers for your product - AI compute is already a commodity, and even if it weren't, you'd have to have commercial terms that put you at risk of not getting paid if the output tokens wind up being worth nothing (fruitless R&D, zero-ROAS campaigns, etc). Which means it makes way more sense to take on that risk yourself. If you believe your tokens are potentially worth more than you can charge for them, then you should be the one building the valuable final products with them so you can capture all that upside. So why isn't the S-1 a detailed plan on how Anthropic will upend the entire digital ecosystem using AI? Why aren't they promising a better Facebook, a better Google, a better Microsoft, a better Steam? They make vague references to being a civilizationally-transformative company, that this is a once-in-a-lifetime investment opportunity, yet they leave the most obvious low-hanging fruit off the table? Unless, of course, AI can't actually do any of these things. Not in an absolute sense - AI can definitely write software, manage infrastructure, run digital marketing campaigns, and all the rest. X is a carnival house of desperate marketing that proves this. But in a relative sense - relative to humans, relative to experience, and relationships, and networks, and trust, and brand. Maybe that's the challenge. Maybe that's why Dario wants a world where Anthropic's customers pay more for tokens that yield greater upside, because the part that yields the greater upside is the geniuses that Anthropic could never put in their datacenter. You know, the geniuses. Us. The people that actually think, and make, and argue, and buy and sell and form opinions and make exchanges and produce what economists classically refer to as "the economy". The ones that generate the never-ending treadmill of shifting taste consensus and regulatory burden and hot-headed competition and races to win new territory and demand for entertainment unlike anything we've seen before. Maybe - and this is a wild concept, but you've read this far - maybe it's the other way around. Maybe Anthropic needs us more than we need them. Even after the massive internet scraping and book burning and tut-tutting. Maybe it takes more than a country of geniuses in a datacenter to build a business that adds value to the economy, and produces products that loyal customers pay for. Maybe - just maybe - Anthropic hasn't quite figured that part out yet.

We’re really just running back every single narrative over the last decade huh

I understand the sky is falling and we’re in the middle of a worldwide everything shortage and everyone is panicking and not even SSL is trustworthy anymore but did it *really* have to break Mr Delivery this evening? Blank menus everywhere 😭

Really loving the stability of Microsoft Defender right now, at this particular moment in time, in the month of October, in the year of our Lord 2026.

Breaking news! The system built inside a computer as a combination of software and data can be precisely modified to alter its behavior in deterministic ways. Probably not the breakthrough they're gonna want to advertise, though.

Just realized what my decades of experience in business systems is actually good for, and it's the exact same skillset that's been winning the day since we put computers to work: Persuading stakeholders NOT to do things 😂

Ah yes, the golden age of AI

There's a BRICS/wall joke waiting to be made here, but for the life of me I can't find it.

Your life is ultimately the balance of your circumstances with your agency. I've seen too many people achieve too many things to believe that it's all a happenstance of birth. But I've also seen too many people give up because circumstances didn't prepare them for the work.

Like $3 trillion invested in AI and it still can't write a hook worth a damn

Sure, I’ll pitch in! The entire notion that online conversations about AI safety are being fed raw and unfiltered right into pre-training is wrong. So we can all relax 😇

It takes 2nd-grader brainpower to compare the vehicle license number showing in the app, to the car you’re climbing into. I don’t know what brainpower it takes to get scammed, but it must be terrifying.