Howdy, it’s Barsee.

Happy Tuesday, AI family, and welcome to another AI Valley edition.

Today’s climb through the Valley reveals:

• Google to launch its first AI glasses in 2026

• Karpathy’s advice: Treat AI as a simulator, not a person

• OpenAI Report: The gap between casual and power users is growing

• You can now order groceries directly inside ChatGPT

• Plus trending AI tools, posts, and resources

Let’s dive into the Valley of AI…

MONGODB

Build in a day. Showcase on a global stage

Join us on January 10th at Shack15 SF for MongoDB’s Agentic Orchestration Hackathon to build and innovate with AI agents. 

• $30,000+ in prizes for the most innovative projects

• Pitch your ideas to top industry leaders and get expert feedback

• Finalists showcase live at MongoDB.local San Francisco on January 15

• Winners are chosen on the mainstage in front of a VIP judge panel of industry leaders

This in-person event is for teams of up to 4, with limited spots. Apply now to secure your spot and see what’s next in AI and data innovation!

_{*This is sponsored}

THROUGH THE VALLEY

1/ Google to launch its first AI glasses in 2026

Google is taking another shot at smart glasses, with a plan to launch in 2026. They are spending $150 million and teaming up with partners like Samsung and Warby Parker to compete with Meta’s Ray-Bans. There will be two versions: an audio-only pair powered by the Gemini assistant, and a more advanced pair with a display for things like maps and translations. Both will run on Android XR. This is a major push for Google as they try to apply lessons from past hardware failures, while competitors like Snap and Alibaba also expand their lineups.

2/ Karpathy’s advice: Treat AI as a simulator, not a person

Andrej Karpathy suggests we are thinking about LLMs the wrong way. He argues that users get better results when they treat the AI as a simulator rather than a being with opinions. If you ask, “What do you think?”, the model just fakes a personality it learned during training. Instead, Karpathy recommends asking how a group of experts would discuss the topic. This framing forces the model to simulate a knowledgeable perspective rather than just chatting, leading to much more useful and grounded answers.

3/ OpenAI Report: The gap between casual and power users is growing

OpenAI’s first report on business usage reveals a massive divide in how companies use AI. Data from over one million business accounts shows that while casual users might send a few messages, "frontier" workers are saving more than 10 hours a week by using advanced features like data analysis. Seventy-five percent of users say AI lets them handle tasks they couldn't do before. The main takeaway is that companies fully committing to these tools are moving faster, while those who don't are starting to fall behind.

4/ You can now order groceries directly inside ChatGPT

OpenAI and Instacart have deepened their partnership, allowing you to go from meal planning to checkout without ever leaving ChatGPT. You can ask for a recipe or help finding ingredients, and the AI will build a cart from local stores and let you pay right in the chat. This is the first time an app has offered full checkout capabilities inside ChatGPT. It’s designed to make the process seamless so users don't have to bounce between apps to get dinner started.

TRENDING TOOLS

• Slop Evader extension > A browser tool that searches the pre-AI internet by returning only content published before November 30, 2022, helping you avoid AI-generated text, images, and videos

• Cofounder > Connect it to Gmail, Slack, Notion, Linear, and automate workflows with plain English commands

• Halftime > Dynamically weaves AI-generated ads into the scenes you’re watching, so breaks feel like part of the story instead of interruptions

• Ripplica > Record any browser workflow once and let context-aware agents replay and adapt it across web apps or legacy systems, even on a schedule, while you are offline

• Orca > An AI engine built for video game development and design

• Gemini 3 Pro > Google’s new vision-driven AI model for document analysis, spatial reasoning, screen understanding, and video intelligence

THINK PIECES / BRAIN BOOST

• One year of using ChatGPT Pro as my first hire

• 10 prompting techniques that experts at Google, OpenAI, and Anthropic use to generate near-perfect accuracy

• The complete guide to Nano Banana Pro: 10 tips for professional asset production

• Has the cost of building software just dropped 90%?

• How Sahil Lavingia runs a $10M company with 1 employee using AI

• Prediction: AI will make formal verification go mainstream 

• We gave 5 LLMs $100K to trade stocks for 8 months

• Chatbots are now rivaling social networks as a core layer of internet infrastructure

•  Google kicks off the Vibe Coding challenge using Gemini 3 Pro in AI Studio

• AI Browsers aren't smart enough yet to take over the internet 

THE VALLEY GEMS

What’s trending on social today:

THAT’S ALL FOR TODAY

Thank you for reading today’s edition. That’s all for today’s issue.

💡 Help me get better and suggest new ideas at [email protected] or @heyBarsee

👍️ New reader? Subscribe here

Thanks for being here.

REACH 100K+ READERS

Acquire new customers and drive revenue by partnering with us

Sponsor AI Valley and reach over 100,000+ entrepreneurs, founders, software engineers, investors, etc.

If you’re interested in sponsoring us, email [email protected] with the subject “AI Valley Ads”.