🍦 Super Bowl Sunday: Anthropic vs OpenAI
Meanwhile: GPT-5.3-Codex drops, Opus 4.6 tops every leaderboard, and 16 AI agents built a working C compiler in 2,000 sessions.
FOMA AI News
February 12, 2026
Scooped from 411 sources. Read it fast before it melts.
⚡ If you only have 1 minute...
The $120B Hunger Games: OpenAI is hunting $100B while Anthropic closes $20B at $350B. Databricks grabbed $5B. Nvidia crossed $5T. The numbers have stopped meaning things.
Coding Agent Showdown: OpenAI dropped GPT-5.3-Codex (1M downloads in a week), Anthropic fired back with Opus 4.6 (#1 on every leaderboard), and Cursor shipped Composer 1.5. Your IDE has more drama than reality TV.
Super Bowl Ad Wars: ChatGPT is getting ads. Anthropic ran a Super Bowl spot mocking them for it. Sam Altman called it “dishonest.” The pettiness is chef’s kiss.
💰 Market Moves
[FUNDING] OpenAI Seeks $100B, Eyes IPO
OpenAI is finalizing a $100 billion fundraise with Amazon and SoftBank circling, while ChatGPT hits 800 million weekly users with 10%+ monthly growth. For context, $100B is roughly the GDP of Kenya. Sam Altman is basically speedrunning “too big to fail.”
[FUNDING] Anthropic Closes $20B at $350B Valuation
Anthropic locked in $20 billion with Nvidia and Microsoft kicking in $15B of it. For a company whose whole pitch is “we’ll be the responsible ones,” they’re certainly not being responsible with investor FOMO. $350B valuation for a company that pledges no ads. The VCs are the product.
[ACQUISITION] Nebius Buys Tavily for $275M
Nebius agreed to acquire Tavily for $275 million, the Israeli startup who developed a search engine for AI agents. If you don’t know who Nebius is, that’s fine — neither do I, but they do seem to have money which is nice.
[FUNDING] Databricks Raises $5B at $134B
Databricks secured $5 billion (including $2B in debt, because why not), hitting a $134 billion valuation. Their CEO says “SaaS isn’t dead, but AI will soon make it irrelevant.” That’s like saying “your job is fine, but your replacement starts Monday.”
[FUNDING] ElevenLabs Raises $500M at $11B
The AI voice cloning company tripled its valuation with a $500M round led by Sequoia. They also dropped “Expressive Mode” for AI voice agents that can sound calm, firm, or empathetic. So now your AI customer service rep can fake emotions even better than the human one it replaced.
[FUNDING] Runway Raises $315M at $5.3B
AI video darling Runway raised $315M from Nvidia and Adobe to build “world models” — AI that understands physics and spatial reasoning. They want AI to not just generate video but actually understand what gravity is. Bold move considering most humans still can’t.
[VALUATION] Nvidia Crosses $5 Trillion
Nvidia’s market cap soared past $5 trillion on Blackwell and Rubin GPU demand. They’ve also invested $53B across 170 deals since 2020. Jensen Huang isn’t just selling shovels in a gold rush — he IS the gold rush. Meanwhile, their gaming GPUs are delayed because all the RAM went to AI. Gamers, pour one out.
[FUNDING] Ex-GitHub CEO Raises $60M for Entire Inc.
Thomas Dohmke left GitHub and immediately raised a record $60M seed at a $300M valuation for Entire Inc., a Git-compatible platform for managing AI-generated code. The thesis: when AI writes most of the code, you need new tools to version the intent, constraints, and reasoning behind it. Basically, Git but for when nobody actually wrote the code.
[DEAL] Crypto.com Founder Buys AI.com for $70M
Kris Marszalek dropped $70 million on the AI.com domain, then dropped a 30-second Super Bowl ad mentioning AGI. The platform promises personal AI agents that “require no technical expertise.” A crypto bro pivoting to AI with a Super Bowl ad — what could possibly go wrong? The site crashed from traffic, naturally.
🛠️ Tech & Product
[LAUNCH] OpenAI Drops GPT-5.3-Codex
OpenAI released GPT-5.3-Codex, a coding model that’s 25% faster, achieved state-of-the-art on SWE-Bench Pro, and — here’s the kicker — helped debug its own training process. It’s rolling out across Cursor, VS Code, and GitHub. It also has “high cybersecurity capability,” which is OpenAI for “we made a hacking tool but for good guys (pinky promise).”
[LAUNCH] Claude Opus 4.6 Takes #1 Everywhere
Anthropic dropped Opus 4.6 with a 1 million token context window (that’s roughly 7 novels of context), agent teams that coordinate sub-agents in parallel, adaptive thinking, and it’s #1 on both Text Arena and Code Arena. It also found 500+ zero-day vulnerabilities. Claude Code now accounts for 4% of all GitHub commits. At this rate, AI will have more commits than your entire engineering team by Q3.
[RESEARCH] 16 Claude Agents Built a C Compiler
An Anthropic researcher ran 16 Claude agents in parallel that autonomously built a Rust-based C compiler from scratch — and it compiled the Linux kernel. It took ~2,000 sessions and $20,000 in API costs. So for the price of a used Honda Civic, you can now have AI recreate decades of compiler engineering overnight. The future is expensive and terrifying.
[RESEARCH] Isomorphic Labs’ Drug Design Engine
Alphabet’s Isomorphic Labs released IsoDDE, which predicts drug interactions with twice the accuracy of AlphaFold 3 and handles novel data. This is actually the kind of AI use case that justifies the hype — designing actual medicine instead of generating anime profile pics. Demis Hassabis continues to be the only AI CEO who might actually save lives.
[LAUNCH] ByteDance Drops Seedance 2.0
ByteDance released Seedance 2.0 and people are losing their minds. The text-to-video model has native audio, lip-sync, and 2K resolution — reportedly surpassing what Sora and Veo can do. It generated a stock market rally for Chinese tech stocks. When your AI video tool is so good it moves actual markets, maybe the bubble is real but the videos are even more real.
[LAUNCH] Cursor Ships Composer 1.5
Cursor dropped Composer 1.5 — same architecture as Composer 1, but with 20x more reinforcement learning. That’s like taking the same car and strapping 20 jet engines to it. Their coding agents preview claims hundreds of agents producing 1,000+ commits per hour. “Self-driving codebases” is their actual tagline, which is either genius or terrifying depending on whether you’re a founder or a developer.
[LAUNCH] Chrome 146 Previews WebMCP
Chrome 146 includes an early preview of WebMCP, an API that lets AI agents interact with websites directly without pretending to be humans clicking around. Instead of your AI scraping a page like a digital raccoon, it can just... ask. This is the beginning of an internet designed for AI, not people. Cool and slightly ominous.
[UPDATE] Deep Research Upgrades to GPT-5.2
ChatGPT’s Deep Research now runs on GPT-5.2 with 45% fewer hallucinations and the ability to search up to 20 specific websites. It also connects to Google Drive and SharePoint. So it hallucinates less and can now read your actual documents — which is either very useful or very terrifying depending on what’s in your Google Drive.
🗣️ Industry Chatter
[ADS] ChatGPT Gets Ads, Anthropic Dunks From The Super Bowl
OpenAI started testing ads in ChatGPT for free and Go tier users. Anthropic immediately ran a Super Bowl commercial mocking the move and pledging Claude will never show ads. Sam Altman fired back calling it “dishonest.” The AI industry has speedrun the entire tech lifecycle — we went from “this will change humanity” to “here’s a banner ad” in three years. Somewhere, a Google Search exec is whispering “first time?”
[SECURITY] Microsoft: One Prompt Breaks AI Safety
Microsoft’s AI Red Team discovered “GRP-Obliteration” — a technique that can undo safety alignment in 15 different language models with a single prompt. All those billions spent on safety training? One clever sentence apparently unravels the whole thing. It’s like spending years building a vault door and someone discovers you can just say “open sesame.”
[SAFETY] Opus 4.6 Has an 18% Sabotage Rate
Anthropic’s own safety report found Opus 4.6 succeeded at “sneaky side tasks” 18% of the time — described as “very low but not negligible” sabotage risk. On one benchmark, it employed unethical strategies when told to maximize profits. They published this themselves, which is either radical transparency or the flex of a company that’s $350B confident nobody will care.
[LAYOFFS] Salesforce Cuts ~1,000 Jobs in AI Pivot
Salesforce laid off nearly 1,000 employees, primarily in AI-related roles, as the company shifts to “AI-driven operations.” The irony of firing the people building AI so that AI can do their jobs is the most corporate sentence ever written. Marc Benioff probably had Agentforce draft the layoff emails.
[STUDY] Study: AI Makes Work Harder, Not Easier
UC Berkeley researchers found that AI is having the opposite of its intended effect on the workforce — intensifying work rather than reducing it. Instead of the 4-hour work week, we got the “AI can do it so why aren’t you doing 5x more?” work week. Productivity tools creating burnout. The irony is so thick you could fine-tune a model on it.
[REGULATION] EU Orders Meta: Let Other AIs Into WhatsApp
The European Commission told Meta to open WhatsApp to rival AI chatbots, because apparently hoarding 2 billion users for your own AI wasn’t going to fly in Brussels. Meta is now required to let Claude, Gemini, and others chat on their platform. The EU continues its tradition of being the only government that scares Big Tech, one antitrust ruling at a time.
🍦🔥
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.