Good morning {{first_name| Artificial intelligence lovers}}. OpenAI has been teasing a super app for months. Today, the first large piece was shipped.
With a major Codex update bringing new features like background computer use, an in-app browser, parallel agents, and more, the rollout is OpenAI’s clearest step yet toward the all-in-one platform it’s created publicly.
-
OpenAI’s super app transformation with Codex update
-
Anthropic’s Opus 4.7 leads the competition and lags behind Mythos
-
Run LLM on your laptop for free with Ollama
-
The first science-specific model from OpenAI
-
4 new tools for AI, community workflows, and more
Obinay
Rundown: OpenAI only Updated Its Codex platform, turning it from a coding proxy into a cohesive ChatGPT + Atlas + Codex application with features like background computer usage, parallel proxies, an in-app browser, image generation, and more.
-
Using your computer in the background allows Codex to run any Mac application on its own, with multiple agents also able to run simultaneously, even in applications without APIs.
-
Memory (in preview) now preserves preferences and context across sessions, while automations allow Codex to pick up long-running tasks days later.
-
The Atlas-powered in-app browser lets developers mark up pages for Codex routing, while built-in gpt-image-1.5 creates mockups without switching applications.
-
Codex has reached 3 million weekly users with 70% month-on-month growth, and Codex President Thibault Sottiaux said OpenAI is “building the super app out in the open.”
Why it matters: Anthropic has had great success with Claude Code and Cowork, and this is OpenAI’s biggest challenge yet – putting Codex on a similar playing field while expanding capabilities beyond just an agent programming assistant. With the company building a “super app,” this seems like the first big shift toward that vision.
Along with beloved
Rundown: Hiring developers or learning to code was the only way to bring the product to life. Lovable completely removes this barrier, as its AI builds real, usable apps and websites from a simple text description.
Millions of users are already creating with Lovable to:
-
Go from idea to customer-ready implementation in minutes, not months
-
Launch your Shopify stores, management tools, landing pages, and more with simple prompts
-
Launch real businesses, validate ideas, and save thousands in development costs
Anthropic
Rundown: Anthropic only Released Claude Opus 4.7, the company’s new top publicly available model that outperforms GPT-5.4 and Gemini 3.1 Pro in proxy encoding – though it still lags behind the company’s unreleased Mythos Preview.
-
Opus 4.7 jumps from 4.6 at 53.4% on SWE-bench Pro to 64.3%, with Mythos Preview still ahead at 77.8%.
-
The model is priced similarly to Opus 4.6 for API usage, although the upgrade uses tokens much faster than its predecessor.
-
Other versions include Claude Code’s default “xhigh” voltage between max and max, and the /ultrareview slash command that flags bugs and design issues.
-
The release comes amid user complaints of performance degradation on 4.6, with early feedback on 4.7 also appearing divided on capabilities despite benchmarks.
Why it matters: Anthropic now runs two parallel tracks: an accelerated two-month public release cadence and a border-gated line in Mythos, accessible only to exclusive partners. This split allows the company to test its strongest models, but it also represents one of the first times the audience feels access beyond real boundaries.
Artificial intelligence training
Rundown: In this guide you will learn how to install Ollama and run a real AI model on your laptop for free. No signup, no account, and no data leaving your device.
-
Go to ollama.com/downloadGet the installer for your Mac/Linux/Windows computer and set it up. Open the app once installed
-
In the app, go to New Chat, select a lightweight model like Gemma3 (about 3GB, suitable for any laptop with 8GB RAM), and wait for it to download
-
After downloading the form, type a claim and press enter. That’s it. You’re chatting with a real AI running entirely on your laptop
-
Try turning on Airplane mode and sending another message to see it work without internet at all
Pro Tip: You can use the Ollama API to give your form access to the web and other tools. You can also point a coding agent like Claude Code to the template and run it for free.
Submitted by Fiddler
Rundown: Evaluating agents with external MBAs seems affordable. So your agent traffic grows. The Fiddler’s Guide explains how to reduce total cost of ownership while eliminating risk gaps in production.
-
Evaluate each effect without sampling
-
Evaluate agents in the environment using in-battery trust models
-
Reduce API costs at scale
Obinay
Rundown: OpenAI Fired GPT-Rosalind, the first model in a new life sciences series designed for drug discovery and biological research, and the company’s first real step into niche thinking in the field – following Tuesday’s GPT-5.4-Cyber.
-
Rosalind can read scientific papers, query laboratory databases, design experiments, and generate biological hypotheses, simplifying the research process.
-
The model demonstrates strong leaps in scientific standards defined for biochemistry, experimental design, instrument use, and more over GPT-5.4.
-
In a blind RNA test conducted by gene therapy lab Dyno Therapeutics, Rosalind’s answers scored better than 95% of human scientists on prediction tasks.
-
The model is available to eligible enterprise users during the testing phase, with companies like Amgen, Moderna, and the Allen Institute already using it.
Why it matters: On Tuesday, it was GPT-5.4-Cyber, today it is GPT-Rosalind. These are two industry models in three days, which shows a trend – the flagship may be good at everything, but the actual huge pegs at the top of industries like network defense or drug design may need purpose-built models.
-
🤖 Cloud Opus 4.7 – The best new AI from Anthropic with advanced proxy coding
-
⚙️ Surfing 2.0 – Agent IDE with new command center, Devin Cloud Agent
-
🚀 Codex Alimentarius – OAI encoding proxy, now with PC use, in-app browser, and more
-
🧠 Hi-World 2.0 – Tencent’s global model that creates interactive 3D scenes
Confusion Rolled A top-tier PC and Mac application that runs agents across 20+ leading models to run native applications, read files, and run Comet Browser 24/7.
Surfing Fired 2.0, Add agent command center with new command center offering for parallel cloud fleets and on-premises agents and bring Devin into the IDE.
Hunyuan Tencent Team Open source HY-World 2.0, is a world model that creates editable 3D scenes with physics-compatible motion, pushing directly into 3D pipelines.
US government And it is said Preparation To grant certain agencies access to Anthropic’s Mythos AI, despite the blacklist and current legal battle with the company.
Alibaba ATH Team foot Happy Oyster is in beta, a new world model that can quickly create interactive 3D environments from multimedia inputs.
In each newsletter we show how the reader is using AI to work smarter, save time or make life easier.
Today’s workflow comes from the reader Jerry J. In Gig Harbor, Washington:
“I’m a 73-year-old author and screenwriter. I’ve been using Claude to build my last two books. Claude helped design the book covers, coordinate the interior design, and even suggest titles for my stories.
Now that it’s in print, Claude is helping me market it to book clubs, libraries, and social media sites. Claude helped me build a group of Substack followers and post new stories every two weeks on Instagram and BlueSky. I am a Claude devotee and will write no more books with his help.
How do you use artificial intelligence? Tell us here.
That’s all 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.
Rowan, Joey, Zack, Shubham and Jennifer – the humans behind The Rundown
