Chrome gets agentic AI upgrade

⋅

Read Online | Sign Up | Advertise

Good morning, {{ first_name | AI enthusiasts }}. 2025 saw new AI browsers popping up left and right… But users didn’t seem too eager to make the switch.

Google is betting they won’t have to, announcing new Gemini upgrades to Chrome that bring agentic browsing and more to a browser that already dominates the market — a default advantage that no amount of AI-first features can easily overcome.

In today’s AI rundown:

  • Google brings more agentic AI to Chrome

  • DeepMind’s science behind AlphaGenome

  • Moltbot (Clawdbot) installation guide & how to use it

  • Two new labs raise big to rethink how AI learns

  • 4 new AI tools, community workflows, and more

LATEST DEVELOPMENTS

GOOGLE

🌐 Google brings more agentic AI to Chrome

Image source: Google

The Rundown: Google just announced a wave of AI upgrades and Gemini integrations into its Chrome browser, including the addition of agentic browsing, built-in image generation, Personal Intelligence, and more.

The details:

  • Auto Browse controls Chrome in its own tab, clicking through sites and completing tasks— also pausing before sensitive actions like payments.

  • Gemini now lives in a persistent sidebar, letting users ask questions, compare products across tabs, and leverage Google apps like Gmail and Calendar.

  • New Nano Banana integration lets users create images directly in the browser, with Personal Intelligence also coming soon for more personalized answers.

Why it matters: There’s been no shortage of AI-first browser competition (OpenAI’s Atlas, Perplexity’s Comet, Dia, etc.) over the past year, but adoption has mostly been lacking — leaving Google with the perfect opportunity to simply continue weaving Gemini into its dominant Chrome platform and integrating deeper with its apps.

TOGETHER WITH HUBSPOT

💰 Turn AI into your income engine

The Rundown: HubSpot’s new “200+ AI-Powered Income Ideas” free guide offers actionable strategies to turn artificial intelligence into your own personal revenue generator — unlocking a gateway to financial innovation in the digital age.

With this guide, you can:

  • Explore hundreds of revenue-generating ideas across industries with real-world applications

  • Follow simple, step-by-step instructions that make AI accessible to everyone

  • Adopt cutting-edge strategies to keep you ahead in today’s fast-paced market

Access your free guide here.

GOOGLE DEEPMIND

🧬 DeepMind’s science behind AlphaGenome

Image source: Google DeepMind

The Rundown: Google DeepMind just published the full research paper and model weights for AlphaGenome, its AI tool unveiled this summer that scans a million letters of genetic code to predict how mutations cause disease.

The details:

  • The model was initially released in June of last year, with the weights and API now freely available for research, and an official paper published in Nature.

  • AlphaGenome was trained on vast genetic datasets and can predict how a single DNA typo affects 11 different biological processes.

  • In tests, it flagged mutations linked to leukemia that sat thousands of letters away from the affected gene, connections that took researchers years to find.

Why it matters: AlphaFold won the Nobel Prize for cracking protein structure, and now Google is taking on the challenge of understanding what 98% of our DNA does. It’s not a “solved” problem like proteins, but giving researchers open access to a map for the genetic code could ramp discovery in ways we’re only beginning to imagine.

AI TRAINING

🦞 Moltbot (Clawdbot) installation guide & how to use it

The Rundown: Learn how to set up your own autonomous agent with starter skills, including image generation and Telegram messaging. After just three days, our Moltbot taught itself to generate images, respond to emails, and deploy its own websites.

Security note: This is experimental, open source code. Do NOT install this on your primary computer. Ideally, you should install Moltbot on a backup computer that’s been wiped clean — you don’t need to buy a Mac Mini, but we don’t want you to lose or expose important documents. Check out this thread for more Moltbot security tips.

Step-by-step:

  1. Grab an API key from your preferred AI model. Grab a free Brave search API.

  2. Run curl -fsSL <https://molt.bot/install.sh> | bash and moltbot onboard --install-daemon in your terminal.

  3. Select “Quickstart,” and follow the prompts.

  4. Your bot will ask you to name it and pick a personality. Now it’s off to the races!

Going further: Watch our full video tutorial to connect your new bot to Telegram, teach it to generate images, and avoid three common beginner mistakes.

PRESENTED BY LESSIE AI

🔎 Your smart people search agent

The Rundown: Tell Lessie AI who you’re trying to reach, and it will explore 100+ premium sources plus the open web to surface verified clients, talent, influencers, and partners.

With Lessie, you can:

  • Search using natural language instead of complicated filters or Boolean queries

  • Surface the most relevant contacts first with intelligent AI ranking

  • Automate personalized outreach messages and follow-ups

Find the right people, for any purpose, in minutes. Use code RUNDOWN for 30% off.

AI STARTUPS

🚀 Two new labs raise big to rethink how AI learns

Image source: Flapping Airplanes / Jerry Tworek

The Rundown: Two AI startups emerged with big investor interest, betting on new AI model paths that diverge from current standards — with Flapping Airplanes raising $180M and ex-OAI’s Jerry Tworek seeking up to $1B for his Core Automation venture.

The details:

  • Flapping Airplanes secured funding at a $1.5B valuation, aiming to train AI that matches human intelligence “without ingesting half the entire internet.”

  • Flapping Airplanes lists Andrej Karpathy and Jeff Dean as advisors, leaning into what Sequoia called the “young person’s AGI lab” approach.

  • Tworek, who left OAI this month, wants to build AI that learns continuously from real-world experience, a capability current systems lack.

  • Tworek plans to focus on a single continual learning model called Ceres, with ambitions ranging from automating factories to eventually terraforming planets.

Why it matters: With top AI researchers openly questioning whether scaling alone can reach AGI, this new wave of startups is a serious contrarian bet against the $200B+ frontier labs. It’s a high-risk play, but investors are pouring billions into the chance that one of these teams cracks the code that the big AI leaders might be missing.

QUICK HITS

🛠️ Trending AI Tools

  • 🧑‍🚀 Agent Composer – AI agents for work that actually is rocket science—compress routine engineering tasks from hours to minutes*

  • 👁️ DeepSeek OCR 2 – SOTA model for reading and extracting text from docs

  • ⚙️ SERA – AI2’s new family of open-source coding agents

  • 🎆 Z-Image – Alibaba Tongyi’s full version of its top open-source image model

* Sponsored Listing

📰 Everything else in AI today

Anthropic is reportedly raising $20B at a $350B valuation, doubling its original target after demand reached 6x what was expected.

Google added Agentic Vision to Gemini 3 Flash, letting the model zoom, annotate, and edit images to answer questions and boosting accuracy 5-10% on visual tasks.

Mistral upgraded its terminal coding agent Vibe to version 2.0, adding custom subagents and workflow skills for developers on its $15-per-month Pro and Team plans.

Anthropic co-founder Jared Kaplan said he believes there is a 50% chance the world’s top theoretical physicists will be “mostly replaced” with AI in the next 3 years.

China reportedly approved ByteDance, Alibaba, and Tencent to purchase over 400k Nvidia H200s, easing a key AI chip bottleneck amid ongoing U.S. tech tensions.

COMMUNITY

🤝 Community AI workflows

Every newsletter, we showcase how a reader is using AI to work smarter, save time, or make life easier.

Today’s workflow comes from reader Nikolaos A. in Clifton, VA:

“I wanted to do renovations to a house I purchased recently. I am building a system of multiple agents, each with a specialty, such as plumbing, concrete/masonry, electrical, etc. These do research on materials and best practices for accomplishing a task.

Additionally, I have an agent that orchestrates the work by routing prompts to one or more agents. Finally, the orchestration agent builds a project book with all the information needed to accomplish the project. I can also upload images that are scanned by a vision model to add to the requirements of a task.”

How do you use AI? Tell us here.

🎓 Highlights: News, Guides & Events

See you soon,

Rowan, Joey, Zach, Shubham, and Jennifer — the humans behind The Rundown

Liked Liked