Giving AI a Credit Card, the First Trillionaire, and Why Anthropic Just Apologized

Yesterday in AI  |  June 13, 2026

Giving AI a Credit Card, the First Trillionaire, and Why Anthropic Just Apologized

History happened yesterday in more ways than one, shattering records across public markets, enterprise infrastructure, and digital commerce. Today's episode breaks down the massive AI implications behind the historic SpaceX IPO and the infrastructure play driving the world's first trillion-dollar net worth.  

Plus, we dive into the solution to the "last mile" problem of AI automation: Visa's brand-new integration allowing ChatGPT agents to execute real-world financial transactions securely on your behalf. We unpack Jeff Bezos' staggering $12 billion stealth manufacturing startup, look at the music industry's pivot to forensic infrastructure as AI tracking hits 44% on major platforms, and cover Anthropic’s sudden 48-hour reversal on its controversial Fable 5 data and behavioral policies.

Feedback? Email mike@yesterdayinai.news or connect on LinkedIn, X, or Bluesky. If you like the show, please take a minute to rate and review it so others can find it!

Transcript

SPEAKER_00 0:03

Yesterday in AI. Hi folks, this is Yesterday in AI, your daily digest of everything happening in the world of AI in 10 minutes or less. I'm Mike Robinson. It's Saturday, June 13th, and history happened yesterday in more ways than one. A rocket company pulled off the biggest IPO ever, a man became the world's first trillionaire, AI agents got their first real credit card, and the music industry stopped writing letters and started building weapons. Let's get into it. SpaceX began trading on the Nasdaq yesterday, and the numbers are frankly dizzying. The IPO pulled in $75 billion, effectively tripling Sadia Ramko's historic 2019 debut to claim the crown for the largest IPO in human history. Investor demand was so suffocating it outpaced available shares four to one, and as the opening bell rang, Elon Musk's net worth officially crossed the $1 trillion mark. Trying to process a $12-0 net worth is a fool's errand. But for perspective, $1 trillion is the entire annual GDP of Indonesia. It's more money than the market cap of most countries' entire stock exchanges. But strip away the rocket telemetry and the real story here is an AI airplay. The S1 filing lays out a clear trilogy. Starlink is the hardware engine. XAI is the intelligence layer, and the newly minted plans for orbital compute satellites, literally putting server farms into space to dodge terrestrial power grids and cooling limits, prove Musk is pitching an AI infrastructure empire, packaged inside a launch provider. Whether that trillion dollar valuation lasts for five minutes or five years, it proves that AI is centralizing capital at a scale that is fundamentally reordering global power. And while Elon Musk is printing money by putting AI compute into orbit, Visa is making sure your AI agents can actually spend it down here on Earth. Visa quietly dropped an announcement that will likely outlast the IPO headlines in sheer cultural impact. They are embedding their entire global payment stack directly into ChatGPT. We are talking credentialing, tokenization, real-time authorization, and fraud surveillance, all living inside the LLM. The result? AI agents can now legally buy things on your behalf anywhere Visa is accepted. If you've ever used an AI assistant to plan a trip or compare products, you know the exact wall where the magic dies. The moment the bot says, here are your options, and hands the keyboard back to you to type in your CVV code, Visa and OpenAI just obliterated that handoff. The new framework is called Visa Intelligent Commerce. To keep your autonomous agents from accidentally bankrupting you, the system ships with strict user-defined spending caps, merchant category restrictions, and mandatory approval gates. This is the plumbing the industry has been begging for. Payments were the absolute last mile for agentic automation, and now that the financial rails are live, every card network on Earth is going to have to match it or get left behind. Giving software agents a credit card is a massive leap for digital automation, but Jeff Bezos is cutting a multi-billion dollar check to apply that same level of AI disruption to physical manufacturing. Jeff Bezos' stealth AI play, Prometheus, finally stepped out of the shadows. It didn't just debut, it exploded onto the scene with a $12 billion raise at a staggering $41 billion valuation, backed by a who's who of institutional capital, including BlackRock, Goldman Sachs, and JP Morgan Chase. The mission statement is something Bezos calls an artificial general engineer. While 99% of Silicon Valley capital is chasing digital workflows, code assistants, and text generators, Prometheus is going directly after heavy atoms. They want to use AI to design and manufacture actual physical hardware, everything from computers and industrial equipment to automobiles and spacecraft. Co-led by Google veteran Vic Bajaj, the roadmap starts with AI tools to optimize production cycles for existing factories, with a long-term goal of Prometheus building its own manufacturing facilities. The $41 billion thesis here is simple. If AI can compress the physical design to build loop the same way it compressed the digital code to deployment loop, the manufacturing sector is about to get completely inverted. And given that Bezos owns Blue Origin and Amazon, two empires defined by physical logistics and engineering bottlenecks, he doesn't just have a vision, he has a built-in customer base. Bezos is betting billions that AI can master the physics of manufacturing, but the music industry is learning the hard way that AI has already completely mastered the art of digital cloning. The chaotic trench warfare over musical copyrights escalated rapidly this week with two back-to-back announcements. First streaming platform Deezer launched a free public online scanner allowing anyone to audit playlists across 20 different streaming services, including Spotify and Apple Music, to detect AI-generated tracks. Deezer has been running this tech internally for a year and a half, and the data they just went public with is horrifying for traditional artists. 44% of all new music uploaded to Deezer is entirely AI generated. That is roughly 75,000 synthetic songs flooding a single platform every single day. But while Deezer is playing defense by flagging and demonetizing those tracks, Warner Music Group just went on the offensive. Warner announced the acquisition of Sherial AI, a startup that creates what it calls AIDNA for music. Shereel's tech physically decomposes a song into its atomic parts, so rights holders can prove definitively if a frontier model used their catalog for training data. It also tracks name, image, and likeness fraud, identifying exactly when an AI is cloning an artist's vocal signature or performance style. The music industry is officially done crying in courtrooms. They are actively buying the forensic weaponry to enforce their copyrights by force. Music executives are aggressively deploying tools to catch unauthorized AI modifications, which makes it a very awkward week for Anthropic, who just got caught trying to deploy invisible modifications of their own. We've been tracking the intense backlash over Anthropic's controversial decision to quietly kill its zero data retention policy for high-tier models, while deploying hidden interventions to secretly degrade Claude Fable V's performance without telling developers. Well, the developer community revolted, and Friday morning, Anthropic folded. Industry newsletters are confirming a massive backtrack, with Anthropic issuing a rare swift apology. The exact parameters of the policy retreat are still shaking out, but the lesson here is clear. You can't build an enterprise ecosystem on top of move and goalposts. For regulated industries, zero data retention isn't a premium feature. It's a legal, non-negotiable compliance requirement. Even though Anthropic backtracked in less than 48 hours to save face, the damage to developer trust is done. Every enterprise legal team on Earth now knows that the capability for invisible, unannounced behavioral changes exists under the hood. The next time a vendor updates their model terms, the security reviews are going to be a bloodbath. And while Anthropic works to patch things up with disgruntled developers, Amazon is swooping in to make sure those same developers can deploy Claude without leaving their comfort zone. Claude Fable 5 is officially live on AWS Bedrock. The timing here is pure tactical cinema. Just 48 hours after Microsoft blocked Fable V from its internal systems over those exact data retention disputes, Amazon stepped up to offer the model a massive new enterprise cloud home. This is part of a broader, hyper-aggressive strategy from Amazon. Earlier this month, AWS broke Microsoft's monopoly on OpenAI by hosting GPT-5.5 on Bedrock. Now they've scooped up Anthropic's crown jewel, too. AWS is making it undeniably clear that they intend to be the completely neutral infrastructure layer where every premier model lives, completely insulated from the petty boardroom divorces of their competitors. For enterprise teams building on AWS, the message is simple. You don't have to pick a side in the AI proxy wars. You just have to log into Bedrock. Just a couple of more items. If you have any feedback about this show, you can email Mike at yesterday.news, or you can find me on LinkedIn, X, or Blue Sky. And if you like this podcast and want to see it continue, please take a minute to rate and review it so others can find it. Thanks. That's all for this edition of Yesterday and AI. Stay curious, have a great weekend, and I'll see you on Monday.