Howdy, it’s Barsee again.

Happy Friday, AI family, and welcome back to AI Valley.

Today’s climb through the Valley reveals:

• Google Gemini turns Chrome into a 24/7 web assistant

• Mark Zuckerberg has begun his quest to kill the smartphone

• OpenAI outperforms humans at the top programming contest

• OpenAI’s future owners revealed

• Luma AI launches Ray3 reasoning video model

• Plus trending AI tools, posts, and resources

Let’s dive into the Valley of AI…

NEBIUS

Switch to Nebius and get more - including up to 3 months off your contract

Bring your AI workloads over, and we’ll cover the migration cost and give you up to 3 months for free. Nebius delivers supercomputer performance with hyperscaler flexibility, so you can build and scale AI models faster.

Learn more & switch today

_{*This is sponsored}

THROUGH THE VALLEY

1/ Google Gemini turns Chrome into a 24/7 web assistant

Google is adding Gemini to Chrome for all U.S. desktop users, bringing conversational search, multi-tab reasoning, and early agent features. The update also packs in new safety tools like scam alerts and one-click password resets.

Gemini can now explain things right on the page through the browser icon, plus summarize or compare info across tabs. A new “AI Mode” in the address bar lets you ask follow-up questions and get page-aware suggestions, rolling out later this month. In the coming months, agent features like cart building, scheduling, and navigation will appear, with users approving the final steps.

Why does it matter?

By turning Chrome into an AI helper, Google is shifting browsing from clicks and searches to delegated tasks. That could reshape shopping, app use, and online navigation, while raising big questions around privacy, transparency, and user control.

2/ Mark Zuckerberg has begun his quest to kill the smartphone

At Connect 2025, Meta revealed new Ray-Ban Display smart glasses and a Neural Band wristband that reads nerve signals (sEMG) so users can silently type in the air. Zuckerberg said he types at about 30 words per minute this way, compared to ~21 wpm for testers and ~36 wpm on phones.

The glasses include a small display for Meta apps, maps, and translation, while still offering cameras, speakers, mics, and built-in AI. The wristband replaces voice commands with silent typing and gestures, like those on Apple Watch or Nintendo Joy-Con. Meta says the combo feels more “present” and less distracting than phones, and it also sidesteps Apple and Google’s app store fees.

Why does it matter?

If typing through a wristband feels natural, computing could shift from phones to wearables. That would reshape daily habits, social interactions, and who controls the next big consumer platform.

3/ OpenAI’s future owners revealed

OpenAI is preparing to drop its capped-profit model and move to a traditional equity structure, valuing its for-profit arm at $500B. Microsoft would own the biggest outside stake at 28% (~$140B), while OpenAI’s nonprofit parent keeps 27% (~$135B). Employees and alumni would hold 25% (~$125B), and a group led by SoftBank, Sequoia, a16z, Founders Fund, and Dragoneer would take 13% (~$65B).

Early backers like Khosla Ventures, University of Michigan, Reid Hoffman’s foundation, Paul Buchheit, and Y Combinator would share about 1% (~$5B). Investors in Jony Ive’s AI startup Io, acquired for $5B in stock, would hold 1.6% (~$7.75B). The deal still needs approval from Microsoft and state regulators, but it would boost investor returns and lock Microsoft’s access to OpenAI’s IP, while giving other backers more value than before.

Why does it matter?

This move marks OpenAI’s shift into a full-scale tech giant. It aligns the company with major investors, secures Microsoft’s position, and ends the capped-profit model that once made OpenAI stand out.

4/ OpenAI outperforms humans at the world's top programming contest

OpenAI’s GPT-5 and Google DeepMind’s Gemini 2.5 Deep Think showed they could beat top human teams at the 2025 International Collegiate Programming Contest (ICPC) World Finals, solving tough algorithm challenges that even stumped some contestants.

GPT-5 scored a perfect run, cracking all 12 problems within the five-hour limit (basically the equivalent of winning gold). Gemini solved 10 out of 12, which would have placed second overall, including one problem no university team managed to solve. Both AI models took part in the official AI track, with answers judged under the same system used for human teams.

Why does it matter?

AI is now not just catching up, but in some cases surpassing the best human teams in tough contests. This shift moves problem-solving from being human-only to a mix of humans and AI, forcing new conversations about education, hiring, research, and what happens once these systems start creating entirely new solutions, not just solving benchmarks.

5/ Luma AI launches Ray3 reasoning video model

Luma AI has launched Ray3, a video model that combines reasoning with cinematic quality output. It can generate native HDR video for professional editing and refine its own results until they meet set standards.

Ray3 also adds annotation tools, letting creators draw directly on frames to guide motion and camera angles. A Draft Mode creates previews in 20 seconds at one-fifth the cost, with final 4K HDR videos delivered in under five minutes.

Why does it matter?

Ray3 is the first reasoning-based video model, moving AI video beyond one-shot generation into iterative refinement. With HDR quality, editing integration, and direct frame control, it brings pro-level video tools closer to everyday creators.

TRENDING TOOLS

• StackOverflow AI > Instant, trusted answers when you need them

• Notion Agents > Automate multi-step workflows, draft documents, and manage databases, running for up to 20 minutes across hundreds of pages

• Meta Ray-Ban Display Glasses > Smart glasses with a private, in-lens color display, powered by Meta’s Neural Band for seamless control

• Luma Ray3 > The first reasoning video model, capable of generating studio-grade HDR content

• Oakley Meta Vanguard > High-performance AI glasses built with tech to push your personal records

• Latitude 2.0 > The first AI agent designed to create other agents

• Thirty Calendar > An AI-powered calendar that organizes both work and life effortlessly

THINK PIECES / BRAIN BOOST

• The Humanoid Thesis

• AI is learning to predict the future and beating humans at it

• OpenAI and Apollo Research released a new study showing that AI models can be sneaky when they think nobody's watching

• AI is learning to predict the future and beating humans at it

• Meta Connect 2025: the 6 biggest announcements

• AI-designed phages

THE VALLEY GEMS

What’s trending on social today:

1/ This prompt will immediately improve the quality of your AI output

2/ The Deepseek-R1 team shows that pure reinforcement learning with answer-only rewards can grow real reasoning skills, with no human step-by-step traces required.

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”.