OpenAI enters browser war with Atlas

PLUS: Ship ideas faster from your browser with Claude Code Web

Good morning, AI enthusiasts. The AI browser war just got its most credible challenger yet, with OpenAI unveiling ‘Atlas’ with agent capabilities, memory features, and deep integrations throughout.

The AI leader’s long-awaited release is here, entering a crowded field with one major advantage — the ChatGPT brand. But is it enough to get users to take the leap away from the comforts of their traditional browsers?

In today’s AI rundown:

  • OpenAI enters browser war with Atlas

  • Anthropic, Amodei respond to government criticism

  • Ship ideas faster with Claude Code Web

  • Origin AI predicts disease risk in embryos

  • 4 new AI tools, community workflows, and more

LATEST DEVELOPMENTS

OPENAI

Image source: OpenAI

The Rundown: OpenAI just rolled out Atlas, a new AI browser that embeds ChatGPT into the web experience — letting the assistant see what users are looking at, remember sites, and leverage agentic capabilities to complete tasks across the internet.

The details:

  • Atlas is “built with ChatGPT at its core,” featuring a sidebar assistant as well as integrations via search, personalizations, and functions like Agent mode.

  • The browser includes optional "memories" that let ChatGPT remember details from visited sites to personalize future responses and automate routine tasks.

  • Agent mode enables ChatGPT to click and complete web tasks autonomously within Atlas, with the functionality initially limited to Plus and Pro subscribers.

  • OAI implemented safety guardrails preventing Atlas from downloading, accessing apps, or taking actions on sensitive sites without explicit permission.

  • Atlas is initially Mac-only in preview for Plus, Pro, and Business tiers, allowing users to import bookmarks, history, and other data from their browsers.

Why it matters: We’ve heard rumblings about OAI’s browser ambitions for a year, and the time has finally come. Atlas immediately threatens Comet, Dia, and other early AI browser entrants, though it doesn’t feel like a release with any groundbreaking functionalities (yet) that will push users to switch from traditional options permanently.

TOGETHER WITH GLEAN

The Rundown: The Glean X AWS Guide to Building AI Agents for the Enterprise is a practical playbook to go from assistants to autonomous agents that handle real work end‑to‑end.

Inside, you’ll learn about:

  • How to define agent instructions

  • Setting guardrails and evaluating performance

  • Managing orchestration at scale

Get the guide to build, govern, and scale agents with confidence.

ANTHROPIC

Image source: Reve / The Rundown

The Rundown: Anthropic CEO Dario Amodei released a statement affirming the company’s commitment to working with the U.S. and “American AI leadership,” coming on the heels of a series of critical comments on X from U.S. AI czar David Sacks.

The details:

  • Amodei highlighted the company’s $200M Department of War contract and widespread federal Claude access to “set the record straight” on policy.

  • The startup also pointed to research showing Claude demonstrating less of a political slant than competing models to push back on claims of bias.

  • Sacks had characterized Anthropic as pushing "regulatory capture" to hurt smaller competition, sparking heated exchanges with investor Reid Hoffman.

  • Sacks also referenced previous comments from Amodei calling Trump a “feudal warlord,” saying Anthropic positioned itself as a “foe” of the administration.

Why it matters: Frontier labs are navigating a delicate balance between staying true to safety principles while maneuvering to secure government contracts and policy influence. With the tech now intertwining with the global power structure, the importance of political positioning within the industry is only going to grow.

AI TRAINING

The Rundown: In this tutorial, you will learn how to use the new Claude Code Web to analyze your codebase, implement features in parallel, and ship changes directly from your browser without switching between tools.

Step-by-step:

  1. Visit claude.ai/code with your Pro/Max account, connect GitHub via OAuth, select "Trusted Network Access", then choose your project repository

  2. Prompt: "Analyze this codebase. Provide: project description, code quality assessment, five feature ideas, and refactoring priorities as a structured report"

  3. Start multiple sessions in the left panel to build features simultaneously - keep them in separate files/modules to avoid conflicts (e.g., "Add profile editing to /components/Profile.jsx" in one, "Implement email notifications in /api" in another)

  4. Review with: "As a senior engineer, review the profile editing feature for code quality, security vulnerabilities, and test gaps. Implement necessary updates"

  5. Click "Create PR" in the chat interface, and Claude generates formatted diffs and descriptions, then merges directly in GitHub without conflicts

Pro tip: Use the "Open in CLI" feature to get a command that connects your local terminal to the Claude Code session. This lets you continue the conversation locally with full IDE integration.

PRESENTED BY POSTMAN

The Rundown: Postman’s State of the API Report reveals critical data on how generative AI, governance, and strategic scaling are transforming the software development lifecycle for engineering leaders.

  • 89% of devs use AI, but few design for agents.

  • 65% of organizations now generate revenue from APIs.

  • 51% cite unauthorized AI agent access as a top security risk

Get the report, including a 90-Day AI Playbook.

AI & GENETICS

Image source: Reve / The Rundown

The Rundown: Nucleus Genomics just released Origin, a suite of AI models that analyze embryonic DNA to predict risks for Alzheimer’s, cancers, and diabetes — also open-sourcing the technology in a first for the IVF industry.

The details:

  • Origin scans 7M genetic markers using algorithms trained on 1.5M people, claiming to reduce disease risks by over 50%.

  • Parents can screen embryos for nine conditions and 2,000 genetic traits with an “IVF+” package starting at $30,000, with partnerships in several major cities.

  • Conditions able to be predicted include prostate and breast cancer, Alzheimer’s, Type 1 and 2 Diabetes, heart diseases, and more.

  • The startup also launched a Genetic Optimization Hub with access to data, models, and metrics, marking the first open-weights system for the sector.

Why it matters: While the price point certainly doesn’t make the screening accessible to the average person, the open-weights release will hopefully allow others to bring AI’s predictive capabilities to a new era of reproductive genetics — though the ethics behind the tech (especially given the price) will continue to be controversial.

QUICK HITS

  • 🌐 Atlas - OpenAI’s new AI-integrated web browser

  • 👁️ DeepSeek OCR - Compress long text for LLMs by treating it as an image

  • ⚙️ Build Mode - New AI-first vibe coding in Google’s AI Studio with Gemini

  • 🎥 Krea Realtime - Create, modify, restyle videos via a near-instant stream

Google AI Studio launched a new vibe coding update, enabling users to build and deploy web apps in minutes via natural language prompts on the platform.

Runway introduced Model Fine-tuning, giving users the new ability to customize the company’s generative video models on their own data and for specific use cases.

Manus released version 1.5 of its AI agent platform, with upgrades including 4x faster task completion, full-stack web development capabilities, and more.

OpenEvidence is raising $200M at a $6B valuation for its “ChatGPT for Doctors” style AI platform trained on medical journals and literature for trusted clinical answers.

Britain’s Channel 4 aired the country’s first TV program hosted by an AI presenter in a segment titled ‘Will AI Take My Job?’, with the use of the tech being revealed at the end to show “how disruptive AI has the potential to be”.

Google launched Skills, a learning platform featuring 3,000 AI and technical courses, with gamified features and employment pathways through company partnerships.

Lovable rolled out a new integration with Shopify, allowing users to build and launch online stores on the platform using natural language prompts.

COMMUNITY

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 Lorna C. in Canada:

"I'm a college professor who decided to drop my old habits of policing and enforcing AI use in my classroom and embrace it. I use Nectir AI and created an AI teaching assistant for my students. I uploaded all my classroom materials, syllabus, and course content, and prompted it to use only my materials (no general knowledge) and not provide any answers, but instead, ask prompting questions. It feels like having an extra teaching partner who will help students learn and not cheat.”

How do you use AI? Tell us here.

That's it for today!

Before you go we’d love to know what you thought of today's newsletter to help us improve The Rundown experience for you.

Login or Subscribe to participate in polls.

See you soon,

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

Reply

or to participate.