In Today’s Issue:
🌌 Google AI Ultra subscribers can now generate and inhabit interactive 3D worlds
📉 Shares plummeted 10% as investors panicked over slowing Azure growth
🎬 Elon Musk’s xAI debuts a unified video-audio generation platform
🧧 ByteDance and Alibaba prepare to launch Qwen 3.5 and Doubao 2.0
✨ And more AI goodness…
Dear Readers,
Google just handed creators the keys to infinite universes—and that's not even the most dramatic headline today. Project Genie is now live for US subscribers, letting you type a few words and walk through photorealistic 3D worlds that generate in real-time around you.
Meanwhile, Anthropic is locked in a $200 million standoff with the Pentagon over where to draw ethical lines in military AI, Microsoft just lost $357 billion in a single day while Meta soared on AI momentum, and ByteDance and Alibaba are racing to drop flagship models right as China celebrates Lunar New Year.
Oh, and xAI's new Grok Imagine API just topped every video generation benchmark out there. Whether you're here for the tech breakthroughs, the market drama, or the philosophical battles shaping our industry—today's issue has something that'll make you think twice about where we're headed.
All the best,
🔥 China’s Big Upcoming Releases
ByteDance and Alibaba Group are gearing up for a high-stakes AI showdown, with both planning flagship model launches around the Lunar New Year in mid-February, a prime window for mass adoption. ByteDance will roll out new language, image, and video models to build on Doubao’s 163M monthly users, while Alibaba counters with Qwen 3.5, optimized for advanced reasoning and tightly integrated across its commerce and payments ecosystem.
📉 Microsoft Suffers Historic Stock Selloff
Microsoft shares plunged nearly 10% after earnings missed key expectations, wiping $357 billion off its market cap and marking its worst daily drop since 2020. Investors flagged slower-than-expected Azure cloud growth (39%), data center capacity constraints, and uncertain returns on AI investments like Microsoft 365 Copilot, even as management defended its long-term strategy. The selloff contrasted sharply with Meta, whose stock jumped 10% on stronger AI-driven results, underscoring diverging investor confidence in Big Tech execution.
🎬 Grok Imagine Redefines Video APIs
xAI has unveiled the Grok Imagine API, a unified video-audio generation and editing platform optimized for quality, speed, and cost. Benchmarks from Artificial Analysis and LMArena rank it #1 overall, outperforming rivals like Sora and Veo on latency-to-quality and price efficiency, making rapid iteration finally practical. With advanced editing, cinematic motion control, and native audio, Grok Imagine positions itself as an end-to-end creative engine for creators, teams, and enterprises.
Tech giants bet on OpenAI comeback
Google Opens the Genie Lamp
The Takeaway
👉 Google launches Project Genie – US AI Ultra subscribers ($250/month) can now generate and explore interactive 3D worlds from text prompts using DeepMind's Genie 3 world model
👉 Sessions capped at 60 seconds – The prototype runs at 720p/24fps with real-time generation; extended session times expected in future updates as computational efficiency improves
👉 Three AI systems working together – Genie 3 simulates worlds, Nano Banana Pro generates visual baselines, and Gemini orchestrates navigation and interactions
👉 AGI implications are significant – DeepMind positions this as critical infrastructure for training AI agents in diverse, realistic scenarios without real-world risks or manual asset creation
Google just handed creators the keys to infinite universes. Project Genie, the company's experimental AI world generator, is now rolling out to Google AI Ultra subscribers in the US. The tool is powered by Genie 3, DeepMind's groundbreaking world model that transforms simple text prompts into explorable, photorealistic 3D environments, rendered in real-time at 24 frames per second.
Here's how it works: You describe a world and a character, and Genie builds it as you move through it. Want to surf massive turquoise waves surrounded by seagulls? Explore ancient Athens in marble glory? Ride a snowmobile through a city sculpted entirely from ice? Just type it in. The AI generates the path ahead based on your actions, maintaining environmental consistency for up to 60 seconds per session.
Project Genie integrates three AI systems: Genie 3 handles the world simulation, Nano Banana Pro sketches your vision into a visual baseline, and Gemini orchestrates your interactions. While current sessions are short and the physics aren't perfect, this is a massive leap from static AI outputs toward truly interactive, AI-generated realities.
DeepMind sees this as more than a creative toy – it's a stepping stone toward AGI. By letting AI agents train in unlimited simulated environments, we're getting closer to machines that can reason and act in the real world. The future of gaming, education, and robotics simulation just got a whole lot more interesting.
Why it matters: Project Genie represents a fundamental shift from AI that generates content to AI that generates entire interactive worlds. This technology could revolutionize how we train AI agents, prototype games, and even how students explore historical eras – all without a single line of traditional game engine code.
Sources:
🔗 https://deepmind.google/models/genie/
I don’t say this lightly: Twin is insane.
You can use Twin to build a fully functional voice agent for a restaurant with just one prompt! It takes reservations, updates a Google Sheet, and sends you a Telegram notification for every call.
It’s incredibly easy, you don't even need to know how to set up Telegram or Vapi. Twin handles both; it connects Telegram in one click and autonomously creates a Vapi voice agent and webhook. You never have to open Vapi at all.
Currently, there's a lot of hype surrounding Clawdbot, and Twin takes this success to the very next level by further developing Clawd's abilities and making them accessible to everyone. An outstanding product!
The AI revolution now makes it possible to set up your own agents using products like Twin, without any coding experience; this truly epitomizes entirely new possibilities.
METR is updating the way they measure model time horizons on software tasks (TH 1.0→1.1). The updated methodology incorporates more of the tasks from HCAST, expanding our total from 170 to 228. This produces tighter estimates, especially at longer horizons.
Anthropic vs. Pentagon: AI Ethics Clash
The AI industry just entered uncharted territory. Anthropic, one of the most safety-focused AI companies on the planet, is currently locked in a high-stakes standoff with the Pentagon over a $200 million contract - and the implications couldn't be more significant for everyone building in this space.
Here's the situation: The Pentagon wants to deploy Anthropic's technology without restrictions, arguing commercial AI should be usable for any lawful military purpose. Anthropic, however, has drawn a line. The company raised concerns that its tools could be used to spy on Americans or assist weapons targeting without sufficient human oversight.
CEO Dario Amodei stated publicly that AI should support national defense in all ways - except those making the U.S. resemble autocratic adversaries. It's the first real test of whether companies building powerful AI systems can maintain ethical boundaries when government contracts are on the line. The outcome will shape how we think about responsible AI deployment for years to come.
Equipment policies break when you hire globally
Deel’s latest policy template on IT Equipment Policies can help HR teams stay organized when handling requests across time zones (and even languages). This free template gives you:
Clear provisioning rules across all countries
Security protocols that prevent compliance gaps
Return processes that actually work remotely
This free equipment provisioning policy will enable you to adjust to any state or country you hire from instead of producing a new policy every time. That means less complexity and more time for greater priorities.
