☕ Yesterday’s AI
Week of October 6, 2025
It’s FOMO either way, so why not just take a cup of coffee and enjoy Yesterday’s News 😊
Welcome to this week’s roundup of AI news, served with a healthy dose of skepticism and maybe too much caffeine. I’ve organized everything into bite-sized sections you can skip to or share with colleagues:
📰 General News - What’s cooking in AI-land
💰 Big Money Deals - Because capitalism never sleeps
🤨 Skeptical Corner - Your weekly reality check
📰 GENERAL NEWS
Meta Launches Business AI
Meta unveiled Business AI, an AI sales agent for small and medium businesses that handles customer interactions across Facebook, Instagram, messaging, and company websites. Also introduced: AI-generated ad music, multilingual dubbing, and virtual try-on features.
Hot take: Meta is basically saying “hey small businesses, you know that expensive customer service team you can’t afford? Here’s a bot instead!” The cynic in me wonders if this is about helping SMBs or just keeping everyone locked into Meta’s ecosystem. Either way, the AI-generated music feature is... interesting. Can’t wait for every Instagram ad to sound like it was scored by the same robot composer. 🎵
Musk Announces Grokipedia
Elon Musk announced xAI is building “Grokipedia,” an AI-powered alternative to Wikipedia, claiming it will be “a massive improvement” and provide “unfiltered truth.” The announcement came after criticism of Wikipedia’s alleged bias.
Hot take: Ah yes, because when I think “unfiltered truth,” I definitely think of the guy who renamed Twitter to X and tried to buy Wikipedia for $1 billion if they’d rename it “Dickipedia.” This is basically “I’ll make my own Wikipedia, with blackjack and my own definition of truth!” The real question: will Grokipedia’s article about Elon Musk be written by Elon Musk about how great Elon Musk is? Place your bets now. 🎰
Google Expands Jules AI Coding Tool
Google expanded its Jules AI coding assistant, an agentic command-line tool that helps developers with coding tasks directly from the terminal.
Hot take: Google releasing yet another tool with a human name (Jules, Gemini, Bard... what’s next, Kevin?). The interesting bit is the “agentic” part - AI that can actually take actions, not just suggest them. Cool in theory, terrifying when it inevitably pushes to production on a Friday at 5 PM.
Microsoft Launches AI Agent Mode
Microsoft announced new AI agent capabilities for its enterprise customers, expanding autonomous task execution across Office 365 and Azure services.
Hot take: Microsoft’s strategy seems to be “sprinkle AI agents on everything and see what sticks.” To be fair, if anyone can make enterprise AI agents boring enough to actually be useful, it’s Microsoft. Your Excel spreadsheet will soon be more autonomous than your intern.
Anthropic’s Claude Codes for 30 Seconds
Reports emerged that Anthropic’s Claude can now write code continuously for up to 30 seconds, showcasing improved sustained reasoning capabilities.
Hot take: “Our AI can think for a whole 30 seconds!” isn’t quite the flex they think it is, but hey, that’s longer attention span than most Twitter threads these days.
Murati’s Thinking Machines Launch
Former OpenAI CTO Mira Murati’s new AI startup, focusing on reasoning models, emerged from stealth mode with initial announcements.
Hot take: The OpenAI talent exodus continues. It’s like watching a really expensive game of musical chairs, except everyone’s starting their own company with nine-figure valuations. The AI bubble definitely isn’t a bubble, guys. Definitely not. 🎈
THOR AI Framework Solves 100-Year Problem
Researchers announced the THOR AI framework solved a century-old mathematical problem in optimization theory, demonstrating AI’s capability in abstract mathematics.
Hot take: Okay, this one’s actually cool. When AI stops generating mediocre marketing copy and starts solving actual hard problems, that’s when it gets interesting. Though I’m sure someone will still find a way to use this breakthrough to generate more clickbait headlines.
💰 BIG MONEY DEALS
Quantum Computing Stocks Surge 3,000%+
Quantum computing stocks went absolutely bonkers this week. Rigetti Computing up 3,700% year-over-year, D-Wave Quantum up 2,500%, IonQ up 600%. The surge followed purchase orders, analyst upgrades, and growing hype around quantum’s potential.
Hot take: Remember when everyone said crypto was volatile? Quantum stocks said “hold my beer.” These companies have essentially zero revenue but multi-billion dollar market caps. It’s like the .com bubble had a baby with the SPAC craze. Some will be Amazon, most will be Pets.com. Choose wisely, or better yet, don’t chase 3,700% rallies. 🎢
OpenAI Acquires Roi Finance App
OpenAI acquired Roi, an AI-powered personal finance app, in an “acqui-hire” where only the CEO joins OpenAI. The app will shut down October 15. This follows similar deals for Context.ai, Crossing Minds, and Alex earlier this year.
Hot take: OpenAI’s strategy: buy small companies, absorb the founder, kill the product, repeat. It’s like they’re Pokemon-collecting talent. The concerning part? They’re clearly building toward a “do everything” AI assistant, and apparently that includes managing your money. What could possibly go wrong with giving an AI chatbot access to your bank account? 😅
Supabase Reaches $5 Billion Valuation
Open-source database platform Supabase hit $5 billion valuation in latest funding round, positioning itself as the open alternative to Google’s Firebase.
Hot take: Nice to see an open-source company actually making money and growing. Though $5 billion for a database service feels very “2025” - everything’s getting valued like it’s going to be the backbone of the AI revolution. Maybe it will be. Or maybe databases are just databases.
Musk Becomes First $500 Billion Person
Elon Musk’s net worth surpassed $500 billion, driven by Tesla, SpaceX ($400B valuation), and xAI ($113B valuation). He’s the first and only member of the half-trillion-dollar club.
Hot take: Half. A. Trillion. Dollars. One person. While we’re all arguing about AI safety and whether chatbots will take our jobs, this guy’s building rockets, cars, brain chips, and a Wikipedia competitor. The wealth inequality is so extreme it’s become parody. Congrats Elon, you won capitalism. Now please fix Twitter.
Former OpenAI/DeepMind Researchers Raise Funds
Another batch of ex-OpenAI and DeepMind researchers announced new AI startups with substantial funding, continuing the brain drain trend.
Hot take: The AI talent merry-go-round continues. Everyone’s raising money, everyone’s going to “build AGI the right way,” and everyone seems to think their approach is the special one. Plot twist: they’re probably all training on the same web scraping data and will arrive at the same place.
Rebellions Raises $250M Series C
Korean AI chip startup Rebellions secured $250 million Series C funding to challenge Nvidia in AI accelerators.
Hot take: “Challenging Nvidia” has become the new “Uber for X” pitch. Everyone wants to be the Nvidia-killer until they realize Nvidia has a decade head start, CUDA lock-in, and basically prints money. But hey, competition is good. Maybe one of these startups will actually ship something. 🤷
🤨 SKEPTICAL CORNER
Your weekly dose of reality to keep us all grounded
Analysts Warn AI Bubble Surpasses Dot-Com
Multiple analysts compared current AI market valuations to the dot-com bubble, noting similar patterns of massive valuations disconnected from revenue. Quantum computing stocks alone saw 3,000%+ gains on essentially zero revenue.
Hot take: Let me get this straight: companies with literally zero revenue are worth billions, quantum computers don’t actually work yet at scale, and everyone’s raising money based on “trust me bro, AGI is coming.” Yeah, this is fine. Totally sustainable. The crash is going to be spectacular.
OpenAI Adds IP Controls After Copyright Issues
OpenAI rolled out new intellectual property controls and removal tools, following mounting pressure over training data copyright concerns.
Hot take: Translation: “We got sued a lot, so here’s a form you can fill out after we’ve already used your content to train our models.” It’s the AI equivalent of asking forgiveness rather than permission, except you’re not really asking for forgiveness either, just providing a complaint box.
Meta to Use AI Chat Data for Personalized Ads
Meta announced they’ll start using conversations with their AI chatbot to improve ad targeting, rolling out later this year.
Hot take: “Hey, tell our friendly AI about your deepest insecurities and financial problems! Oh by the way, we’re going to use that to sell you stuff.” This is either brilliant monetization or a dystopian nightmare, depending on which side of the transaction you’re on. Probably both. 😬
Wiz CTO Warns AI Fueling Cyberattack Surge
Cloud security company Wiz’s CTO warned that AI is dramatically increasing both the volume and sophistication of cyberattacks, with AI-generated phishing and malware becoming mainstream.
Hot take: Turns out giving everyone access to powerful AI tools means the bad guys get them too. Who could have seen this coming? (Everyone. Everyone saw this coming.) AI-generated phishing emails that are actually well-written and convincing? Your grandma doesn’t stand a chance. Hell, I don’t stand a chance.
🎬 That’s a Wrap!
This week in AI: quantum stocks defy physics (and economics), everyone’s building Wikipedia clones, and the robots are coming for your customer service jobs first.
See you next week, assuming we haven’t all been replaced by AI agents who write better newsletters.
YAI 👋
Found this useful? Share it with a colleague who needs a reality check on AI hype!
Have thoughts? Reply and let me know what you think - I actually read them (unlike the AI) 😉



This was the most entertaining news post about AI that I’ve read 😂 thanks for making news more fun. Really enjoyed your hot takes.
oooh thanks so much ♥️