Amazon's AI panic mode activated

Amazon just handed 27-year veteran Peter DeSantis the keys to their AI kingdom while Rohit Prasad (Alexa's dad) mysteriously "explores new opportunities" 👀
They're throwing $8 BILLION at Anthropic and eyeing $10 billion for OpenAI because apparently that's how you fix being last place in the AI race??
Amazon's having its "oh shit we're actually behind" moment and trying to buy their way to relevance. Classic. (source)

Google's "TorchTPU" project is basically trying to make their chips speak PyTorch fluently - with Meta's help because of course these frenemies are teaming up
They're targeting Nvidia's biggest moat: that CUDA software ecosystem that's more embedded in AI than JavaScript is in web dev (and that's saying something)
Google's trying to break Nvidia's stranglehold on AI computing but like... good luck dethroning the GPU gods bestie 💀 (source)

Chinese scientists built a working EUV lithography prototype in Shenzhen using ex-ASML engineers with fake names (not suspicious AT ALL)
The machine is so big it fills an ENTIRE FACTORY FLOOR - targeting 2028 for actual chip production
China might be wayyy closer to semiconductor independence than anyone thought. The sanctions are... not sanctioning?? (source)

Meta's AI chief is raising 500 million euros at a 3 BILLION EURO valuation for a startup that doesn't exist yet
His big idea?? AI that understands the physical world using "world models" (revolutionary concept: AI that knows things fall down)
Either this is peak bubble behavior or LeCun knows something we don't. Probably both tbh. (source)

Tech stocks TANKED Wednesday - Nasdaq down 1.81% as investors suddenly realize AI spending might be... unsustainable??
Oracle dropped 5.4% after their data center partner ghosted them on a $10B deal; Nvidia fell 3.8% (still worth more than most countries tho)
The market's asking the trillion-dollar question nobody wants to answer: When does all this GPU hoarding actually make money? (source)

New model supposedly outperforms GPT-5.2 (which doesn't exist but ok) while being 3x faster than before
Google's processing 1 TRILLION tokens daily on their API because apparently that's a normal Tuesday now
The AI wars are so intense we're getting weekly model drops like it's fast fashion but for chatbots. (source)

The FTC is probing Instacart's Eversight AI after it made some shoppers pay 23% MORE for the same groceries (not very demure, not very mindful)
Shares dropped 10% because turns out using AI to squeeze customers isn't a great look??
This could set huge precedents for AI pricing. Spoiler: "the algorithm did it" isn't a legal defense. (source)

Author Elizabeth Lyon claims Adobe used the infamous "Books3" dataset" (191,000 pirated books) to train SlimLM
This is like the 47th lawsuit over this same dataset but who's counting
The copyright wars continue and nobody knows if training on pirated content is illegal or just ethically questionable :/ (source)

Anthony Enzor-DeMeo promises Firefox AI will be completely optional (revolutionary concept in 2025)
Mozilla trying to escape Google's search money while competing with AI-first browsers
Finally someone offering a middle path between "AI EVERYTHING" and "no AI ever." We love to see it. (source)

You can now build mini apps in Gemini using natural language because apparently typing "make me an app" is programming now
The visual editor lets you create apps without code (your CS degree is crying rn)
No-code development is eating the world and junior devs are sweating PROFUSELY. (source)

New AI can isolate ANY sound from audio/video using text prompts
Works for music, speech, and that one guy coughing during the quiet part
Could revolutionize hearing aids and music production... or just make deepfakes even more terrifying!! (source)

Skana Robotics cracked underwater robot communication using old-school math AI instead of hallucination-prone LLMs
Because when you're defending submarines, you want predictable AI, not creative writing
Solving underwater comms could transform naval defense. Also robots talking underwater is lowkey terrifying. (source)

Top researchers share their 2026 AI predictions ranging from "cautiously optimistic" to "we're all doomed"
Topics include model capabilities, regulations, and whether we'll all have jobs
Academic perspectives are the only grounded takes left in this hype tornado. (source)