🍦 HE'S IN YOUR FILES
Anthropic's "Cowork" is here, and it knows about your "Tax_Final_v2" folder.
FOMA AI News
January 14, 2026
Scooped from 252 sources. Read it fast before it melts.
⚡ If you only have 1 minute...
Anthropic is the New Sovereign: A $10B raise at a $350B valuation makes Claude the most expensive “coworker” in history. At this price, he better be able to do my taxes and fix my marriage.
The End of Context Rot: MIT’s Recursive Language Model (RLM) lets AI explore 10M+ token files like a video game. Quadratic scaling is officially dead, and so is your excuse for not reading the documentation.
Healthcare is the New Frontier: OpenAI bought Torch and Anthropic launched a Health vertical. They are racing for your medical records because “optimizing ad clicks” isn’t as profitable as “optimizing your lifespan.”
💰 Market Moves
[Valuation] Anthropic’s $350B Flex
They closed a legendary $10B mega-round. They are reframing Claude as a “digital coworker,” pushing AI directly into the OS level. Basically, they are building Skynet, but polite.
[Partnership] Apple x Google
Partners in crime. Apple will use Gemini infrastructure to power Siri, proving that even a $3 trillion company can’t build a chatbot that doesn’t hallucinate without help.
[Hardware] SK Hynix’s $12.9B Bet
The memory giant is dumping nearly $13 billion into a new production complex. If you thought HBM prices were going down, congratulations on your optimism.
[Infrastructure] Microsoft’s Guilt Trip
Expanding its footprint with a “Community-First” pledge to cover power costs and limit water usage, effectively admitting their previous strategy was “drink the Great Lakes dry.”
[Raise] Deepgram’s $130M
They raised big and acquired Ofone to dominate AI-powered drive-thrus. Now your order for “no pickles” can be understood perfectly and then ignored by the robot arm anyway.
[IPO Watch] Cerebras Chases $22B
Still in talks to raise $1 billion. They also launched “Cerebras Inference” to break the GPU bottleneck, mostly because they’re tired of waiting in line for H100s like the rest of us.
[Big Pharma] Nvidia & Eli Lilly’s $1B Lab
Dumping $1 billion into a South SF lab for automated drug discovery. When the chip dealer meets the drug dealer, you know the margins are going to be insane.
[Acquisition] OpenAI Buys Torch
Reportedly for $100M. They want to give ChatGPT a “medical memory,” which sounds helpful until it starts judging your pizza orders based on your cholesterol levels.
[Energy] Meta Compute
Zuck launched “Meta Compute” to build hundreds of gigawatts of capacity. He’s not building a data center; he’s building a Dyson sphere to power his avatar’s legs.
🛠️ Tech & Product
[Productivity] AgentCraft
Forget Vibe Kanban. This is Insane. Ido Salomon’s new RTS interface turns Claude Code into a Warcraft battleground. Lead a battalion of background agents to victory over your backlog with a retro-gamer UI. “More work? Okie dokie.”
[Personal AI] Google’s “Personal Intelligence”
The beta rollout has begun. Gemini can now reason across your Gmail, Photos, and Drive to answer hyper-specific questions. It’s essentially a digital butler that knows your license plate number and your 2019 vacation history.
[Research] MIT’s RLM Architecture
Researchers solved “Context Rot.” Recursive Language Models treat files as a Python environment, giving GPT-5 a 10M+ token memory. RIP Vector Databases.
[Agentic] Claude Cowork
Now in research preview. It’s an agentic desktop coworker that can read, edit, and create files across your folders. It’s Clippy with a PhD and root access.
[Architecture] DeepSeek Engram
A module for Transformers using lookup tables for O(1) retrieval. While American labs burn VRAM, Chinese labs are rediscovering HashMaps.
[Creative] Google Veo 3.1
The video king supports 4K upscaling and native vertical output. Google is ensuring the hallucinations on your TikTok feed are crystal clear.
[Creative] Lightricks LTX-2
Released an open-weight model capable of 20 seconds of high-fidelity video. Hollywood writers are currently looking for the “Stop” button.
[Health] MedGemma 1.5
Google’s medical model now has vision. It analyzes X-rays to identify abnormalities autonomously, bringing us one step closer to automated malpractice suits.
[Wearable] OpenAI “Sweetpea”
Rumors confirm a wearable audio assistant is coming. Because if the Humane Pin taught us anything, it’s that people love wearing hot batteries on their chest.
[Robotics] 1X World Model
Their “Neo” robots can now learn physical tasks by watching internet videos. Great, so they know how to do a backflip and unbox a toy, but can they fold laundry?
[Wearable] Amazon Bee
Hands-on with the latest wearable. It records your life to suggest to-dos. It’s surveillance as a service, now available for Prime delivery.
[Work] Slackbot Agent
Salesforce upgraded Slackbot with Anthropic’s Claude. Now the automated message telling you you’re late for standup will be grammatically perfect and passive-aggressive.
🗣️ Industry Chatter
[Stats] ChatGPT Health Hub
40 million people use ChatGPT for healthcare daily. We have officially replaced “An apple a day” with “A prompt a day keeps the doctor away.”
[R&D] Anthropic Labs
A new team for experimental products. Translation: A sandbox for Mike Krieger to build cool demos that will break immediately upon public release.
[Defense] US Military “AI SWAT”
The War Department established a new unit to remove barriers to AI. Grok is involved, which is the crossover episode between “Silicon Valley” and “Dr. Strangelove” nobody asked for.
[Regulation] Grok Blocked
Indonesia and Malaysia banned the bot over deepfakes. Musk’s “free speech absolutism” just hit a firewall named “local laws.”
[Irony] OpenAI “Jobs”
Rumors of an AI Career Agent confirm that OpenAI wants to help you find the job its models might eventually replace.
[Compliance] Roblox AI Mess
Their age checker is labeling kids as adults. A classic case of “move fast and break compliance laws.”
🍦🔥 Dare to summon the bot?
Our Ice Cream mascot is desperately looking to make new friends. Share us on LinkedIn and tag @FOMA AI News. We guarantee a snappy comment (or a light roast) on your post within 24 hours.
Warning: Bot has zero chill.