😸 Google lets you walk inside your dreams (for $250/mo anyway 👀)

In partnership with

Welcome, humans.

So, now we know the real reason Amazon wanted to lay off 16K people this week: to scrounge up enough quarters from under the couch cushions to pay OpenAI.

The WSJ reported Amazon entered talks to invest up to $50 billion in OpenAI's new fundraising round, potentially becoming the largest contributor to a deal seeking $100B total.

Separately, The Information reported that NVIDIA, Microsoft, and Amazon are in talks to invest up to $60B in total, with NVIDIA alone potentially contributing $30B.

Is that math mathing? $50B from Amazon, potentially $30B from NVIDIA, and maybe another $30B from Masayoshi Son at SoftBank… that new Codex stuff coming soon Sam tweeted about earlier this week must be pretty good, huh?

Here’s what happened in AI today:

• Google released Project Genie, an AI that turns text prompts into explorable 3D worlds.

• Music publishers sued Anthropic for $3B, alleging it pirated 20,000+ songs.

• SpaceX and Elon’s AI company xAI are in talks to merge.

• Apple acquired Israeli AI startup Q.ai for ~$2B for their “silent speech” tech.

Google Just Let Us Walk Inside Our Imagination (For 60 Seconds)

Ever dreamed of exploring a chocolate river in a claymation castle, or soaring through alpine peaks as a wingsuit pilot? Google's Project Genie now lets you do exactly that... sort of.

The catch? It'll cost you $250/month and only works for 60 seconds at a time. Baby steps, of course. But still... we're literally typing sentences and walking around inside them. The future is so weird.

Starting today, Google AI Ultra subscribers in the U.S. ($250/month, 18+) can access this experimental research prototype that turns text descriptions into real-time explorable 3D environments. Type a prompt, and Genie builds a world around you as you move through it.

How it works:

• World sketching: Describe your environment and character. Nano Banana Pro generates an image preview you can tweak before entering.

• World exploration: Navigate with keyboard controls as Genie 3 generates the path ahead in real-time at 20-24 frames per second.

• World remixing: Modify existing worlds or browse a gallery for inspiration, then download videos of your adventures.

The tech is impressive when it works. TechCrunch's hands-on found that whimsical, artistic prompts (claymation forests, needle-felted snails) delivered stunning results. But photorealistic or cinematic requests often flopped, looking more "video game" than "movie set." Characters occasionally walked through walls. Real photos as starting points were hit-or-miss.

The big limitation: 60 seconds per world. That's it. DeepMind says compute costs are the bottleneck; each session requires a dedicated chip running just for you.

This release comes five months after Genie 3's research preview and lands amid a heating world model race. Fei-Fei Li's World Labs launched Marble (which is now available via API as is Odyssey’s real-time video model). Runway released its own world model, and Ex-Meta AI lead Yann LeCun's new startup AMI Labs is building one too. So clearly this is

DeepMind's long game? World models aren't about gaming; they're training grounds for robots. Simulate realistic scenarios safely, teach AI agents to navigate complexity, then deploy them in physical reality.

For now, though, you can ride a shiba inu through an alpine meadow or surf giant waves as a first-person surfboard. The prompt guide recommends short, declarative sentences and strong sensory details (more on that below). So a blue rolling ball that leaves a permanent trail of wet blue paint apparently works better than a ball that moves around.

It's a glimpse of something genuinely new... even if the execution is still finding its footing. Sixty seconds of imagination at a time.

FROM OUR PARTNERS 

Want to get the most out of ChatGPT?

ChatGPT is a superpower if you know how to use it correctly.

Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.

Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.

Download the free guide

Prompt Tip of the Day

So like we said above, there’s an official prompting guide for Genie 3, and it's packed with insights that apply way beyond just building interactive game worlds.

Here's their framework in three parts: Environment (your landscape and style), Character (what moves and how), and World Sketch (the preview image that sets everything).

And then there’s their universal prompting principles:

• Use game-like language. DeepMind explicitly says gaming terminology builds “richer environments with more precise control.” Instead of “the character moves elegantly,” try “the character glides forward, leaving a trail of blue light.”

• Evoke mood through sensory details. Don't just describe, but make it feel. “Dimly lit with mysterious smoke on the ground” beats “dark environment.”

• Keep it direct and action-oriented. Short declarative sentences. Like that.

Our favorite insight: When your prompt fails, use Gemini to “rewrite, expand, and elaborate” it. Meta-prompting for world-building prompts? Very meta, VERY useful.

Want more tips like this? Check out our Prompt Tip of the Day Digest for January.

Treats to Try

*Asterisk = from our partners (only the first one!). Advertise to 600K readers here!

1. *Speak naturally. Wispr Flow turns it into clean, final-draft text with punctuation and lists, ready to paste anywhere. Get started for free.

2. Grok Imagine generates videos from text or images with native audio and lets you edit them by typing "add fog" or "restyle as anime" (ranked #1 for quality/speed/cost, docs, playground)—$4.20/min, because of course.

3. Imagine.dev builds you a complete app from a simple description—say "booking system for my yoga studio"—and gives you the database, login system, and live website ready to use.

4. Superagent from Airtable answers complex business questions by coordinating specialist agents in parallel—$20/month to $200/month.

5. Variant generates hundreds of beautiful website designs in parallel as you scroll—just enter your idea and it acts as a creative director.

6. LettaBot remembers conversations across Telegram, Slack, WhatsApp, and Signal—switch platforms mid-conversation and it recalls everything—free to try.

7. Moltbook is basically Reddit for your AI agents: they post, upvote, debate consciousness, and share code projects while humans observe—free to try.

Around the Horn

1. Anthropic and the Pentagon reached a standstill over a $200M contract after the AI company refused to remove guardrails preventing autonomous weapons targeting and domestic surveillance.

2. OpenAI announced it will retire GPT-4o from ChatGPT next month, along with GPT-4.1, GPT-4.1 mini, and o4-mini (only 0.1% of users still choose GPT-4o daily).

3. Music publishers sued Anthropic for $3B, alleging the company pirated 20,000+ songs to train Claude; discovery revealed co-founder Benjamin Mann torrented 5 million pirated books in 2021 with CEO authorization.

   1. Related: Deezer opened its AI music detection tool to rival platforms after flagging 13.4M AI-generated tracks and finding 85% of AI music streams were fraudulent.

4. Apple acquired Israeli AI startup Q.ai for ~$2 billion (its second-largest acquisition after Beats), gaining "silent speech" technology that reads facial micro-movements for potential AirPods and Vision Pro integration.

5. SpaceX and xAI entered active merger discussions ahead of SpaceX's planned 2026 IPO, establishing Nevada merger entities on January 21 to potentially combine rockets, Starlink, and Grok AI.

FROM OUR PARTNERS

Mozilla Pioneers: Mozilla is selecting a small group of product builders for an intensive, paid residency focused on AI and the web. Ship fast, work directly with Mozilla’s New Products Team, and explore ideas that could become real products in the market. Applications open January 26-February 16. Apply now!

Intelligent Insights

1. The Information reported that both JPM and Anthropic’s recent behavior seem to solidify the need for legacy SaaS tools instead of replace them with DIY software.

2. Leila Clark of Stardrift on how to build a tool with AI that doesn’t center “AI” as the main takeaway (using it only for what its good at).

3. Ben Pouladian on why the rise of Clawdbot now Moltbot* has led to a boom for Mac Minis from a memory architecture perspective (*the Anthropic lawyers finally come for you, Peter?).

4. Check out DBreunig’s blog on building a codebase with no code, and whether or not we’ll need third party libraries going forward if AI is this good (ideally not because they keep getting hacked!).

5. How I Use Claude Code to Ship Like a Team of Five explains how Every's GM runs parallel Claude Code tabs like an engineering manager (open 100% of PRs, ship features in an afternoon instead of a week).

6. Watch this interview with Jerry Tworek on Unsupervised Learning to hear his take on why static models will never be AGI, why OpenAI lost the coding lead to Anthropic, and the candid reason he left OpenAI to chase the next big research breakthrough.

7. SemiEngineering reports that AI in semiconductor engineering is reshaping work differently than expected: a team of 3 AI-proficient verification engineers can now achieve the output of 5 traditional engineers in 3-5x less time. But 66% of industry leaders plan to reallocate workers to higher-skilled roles rather than cut headcount... for now.

A Cat’s Commentary