Apr 2 • 🫖 AI News - The Tea On AI
What's tea in AI? 🫖You're Weekly AI News Update 4.1
The Tea On AI:
🍵 MAIN CHARACTER ENERGY: ANTHROPIC LEAKED ITS OWN SECRET WEAPON
Ok so they accidentally dropped the bag. ASecurity researchers found nearly 3,000 unpublished Anthropic documents sitting unprotected in a database — including draft blog posts for a new model codenamed Claude Mythos .
Here's the tea: Mythos sits above Opus — Anthropic's own draft literally said it's "larger and more intelligent than our Opus models, which were, until now, our most powerful." It scores "dramatically higher" on coding, reasoning, and cybersecurity. That's talk Anthropic has never used before.
The real plot twist? Internal notes say Mythos "could outpace defenders" in cyber capabilities aka be autonomous , making it more capable of taking actions without you and doing things without your approval. So Anthropic is essentially scared of their own model. That's not a product launch, that's a warning shot.
🧠 Why it matters to YOU: Premium AI is about to get priced like enterprise software. If you're not thinking about the cost of tiered AI access, start now.
⚡ QUICK HITS — HERE'S WHAT ELSE YOU MISSED
THE COURTS SAID NOPE
A federal judge blocked the Trump administration from banning Anthropic's federal contracts the Pentagon had called Claude a "supply chain risk." The judge cited First Amendment retaliation and granted a 7-day appeal window. Case is still live in San Francisco, but Anthropic got the W , for now. The bigger story? Government AI contracts are now a legal battleground, and whoever wins federal access wins decades of institutional lock-in.
🍵 APPLE WAVED THE WHITE FLAG (KINDA)
Apple announced iOS 27 will let Gemini, Claude, and other AI assistants compete directly inside the iPhone through a new Extensions system. Meanwhile, Apple is secretly building its own chatbot codenamed "Campos" powered by Google's Gemini. They also poached Google's Lilian Rincon as VP of AI product marketing ahead of WWDC on June 8.
Translation: Apple stopped pretending it could win the AI race (because they don't got it without Steve Jobs) and opened the iPhone to 1B+ users as a distribution channel for every AI company. That's a massive distribution event.
🍵 RIP SORA
OpenAI shut down Sora its video generation app and API just six months after launch, taking a reported $1B Disney content deal with it. They're allegedly reallocating compute toward a new model codenamed "Spud." Oh, and Sam Altman quietly stepped back from safety oversight to focus on data center expansion and fundraising. The move: OpenAI is in a full sprint and safety is not the finish line.
🍵MICROSOFT AND OPENAI ARE ON A BREAK
Microsoft stepped in to take over a large Texas AI data center project that OpenAI declined to pursue. And earlier this week, Medium's top story called it outright: "The Microsoft-OpenAI Divorce Is Official: Inside the $250 Billion Betrayal." This is big for corproate and enterprise customers who built on Azure OpenAI integrations should be paying attention.
🧠 GOOGLE SAID "YOU CAN'T LEAVE ME"
Google had its biggest AI day of the year — dropping Gemini 3.1 Flash Live, expanding Search Live to 200+ countries, and launching the pettiest feature of 2026: chat history import from rival AI apps. You can now drag your ChatGPT or Claude conversation history straight into Gemini via copy-paste or ZIP upload. This is the retention play of the year. Google isn't building a better model , it's building a better ecosystem out of your context.
I'll keep saying this, but Google is going to win the AI race and be the top player in the game!
🍵 POLICY DRAMA: THE TEA FROM DC
The political layer on AI is getting messy messy.
Bloomberg's headline said it plainly: "AI Schism Grips Washington as Tech, Labor Vie for Upper Hand. Silicon Valley and Trump officials are increasingly at odds over governance, labor impacts, and which companies get federal contracts.
Bernie Sanders proposed a moratorium on new data centers. The Washington Post called it "his dumbest idea."
Defense News warned that the military's "human in the loop" framing for AI decisions is "dangerously misleading" suggesting the loop is often ceremonial at best.
America's first AI-fueled war is reportedly active, per NPR, with AI playing operational roles in conflict in Iran. This is no longer theoretical.
Switzerland 2027 is emerging as the venue for a global AI treaty, Ambassador Thomas Schneider is leading diplomatic efforts toward an international governance framework.
And for the people thinking about economic policy: AI economist Alap Shah proposed an "American Prosperity Compact" — a four-tier economic response to AI labor displacement, including an AI Dividend Fund modeled on Alaska's Permanent Fund. This is the policy blueprint people will be citing for years.
🍵 THE DROP: NEW TOOLS THAT SLAPPED THIS WEEK OR ARE COMING SOON
🔧 Tool What It Does
- Claude Mythos - Most capable AI built to date — still in controlled testing ( not vailable yet)
- Suno v5.5 | Clones your verified singing voice; 2M paid subscribers, $300M ARR
- Granola Meeting notetaker turned enterprise AI app raised $125M at $1.5B valuation
🍵 THE WORKPLACE WATCH: AI AT YOUR JOB
Your AI use is officially a performance metric. Built In ran the feature: "Your AI Use Is a Performance Metric Now. Here's How to Talk About It." Organizations have shifted from "do you use AI?" to evaluating how effectively you do.
The "AI Enablement" job title is having a moment. LinkedIn surfaced openings this week from Dragos, Volaris Group, Cvent, Blackbaud, Exact Sciences, and AILY LABS all hiring for AI Enablement leads and AI literacy program managers. This role category didn't exist at scale 18 months ago.
AI is also creating jobs in the trades. CIO Dive reported that AI is spurring demand for skilled trade workers, construction and infrastructure jobs are growing in response to data center buildout.
📌 Sources this week: The Neuron, CIO Dive, The Prohuman AI, Nextool AI, The Tech Buzz, Built In, Every, Defense News, NPR, AP News, Bloomberg, Politico, Nature Medicine — all synthesized from Jeneba Wint's inbox · Powered by Claude
Tell me what you're paying attention to this week 👇🏾
1
6 comments
What's tea in AI? 🫖You're Weekly AI News Update 4.1
powered by
Suggested communities
Powered by