🏠 OpenAI is going BIG

GM! It’s Brett.

Been a minute! I promise I hibernated from writing for a good reason. I’ll share some time next month.

Until then, here are your stories for the week!

Brett

🔥 OpenAI is going BIG

On November 4, OpenAI announced plans to acquire Pierre, the company behind Code Storage and developer tools that help people navigate files and review code changes. Two of its libraries, Trees and Diffs, are already inside Codex.

The deal hasn’t closed. Pierre’s open-source projects will stay open-source, and OpenAI says Code Storage customers can keep using the service after closing. The announcement.

My read: OpenAI wants more of the journey from “I have an idea” to “customers are using it” to happen inside Codex. Generating the code is one part. Understanding it, reviewing it, and getting it into production are where a lot of the work lives.

There’s a pretty wild example from inside OpenAI itself.

On November 2, the company published a breakdown of Pando, its new Git backend. It says a single engineer designed the core, teams of agents implemented it, and a handful of engineers deployed it. Kickoff to production took ten weeks.

Pando now powers development in OpenAI’s monorepo. The company reports nearly five times lower median push latency and almost a petabyte of Git data served daily. These are OpenAI’s own reported results, but they’re a much more interesting example than another agent building a to-do app. Read the engineering breakdown.

The founder takeaway: when you can try more ideas, choosing which ones deserve to exist becomes a bigger part of the job.

A working prototype gets you into the conversation. You still need a point of view, a product people want, and a way to reach them.

Which brings me to the role I’ve been writing about.

In partnership with Framer

Design, build, and update your site with ChatGPT, Claude, Cursor, and more using Framer MCP. All the power of AI, with the flexibility and control of Framer.

Explore Framer MCP.

💥 Every company wants a head of storytelling and design

I posted about this in October because companies kept hitting me up looking for the same kind of person.

Someone who has taste, understands AI, and can actually make things: brands, landing pages, product experiences, videos, images.

They can spot what’s interesting about a product and turn it into something people want to look at. Then they can get it in front of the right people.

That combination is incredibly valuable when everyone has access to better building tools.

You can generate five landing pages. You still have to know which one makes your customer understand the product. You can make a launch video. You still need an idea worth watching.

And you need evidence that you can do it. Put your designs and products into the world. Publish the work. Learn what people respond to.

The hiring argument in my post was deliberately strong. But the advice applies even if you never want “storytelling” in your title: develop taste, learn to build, and get comfortable sharing what you make.

Read my original post.

🧠 ON MY MIND

ChatGPT’s image tools got another upgrade. On November 4, OpenAI announced Images 2.6, with faster generation and better preservation of details across edits. GIF creation is rolling out, and saved references let you reuse an image in later chats. Useful if you’re trying to make a character show up consistently across a campaign. Release notes.

Always-on agents are reaching more workplaces. Dots became generally available to eligible Enterprise, Edu, and Healthcare workspaces on November 4, with an admin preview through November 18. They can keep working between conversations. Admins get controls over access, connections, and activity. The rollout still has eligibility and residency restrictions. Details.

Computer use is getting cheaper. OpenAI’s October 26 GPT-6.1 Terra announcement puts standard API pricing at $1 per million input tokens and $5 per million output tokens, half the standard prices of GPT-5.6 Terra and GPT-6.1 Sol. The practical question for founders is which repetitive tasks become worth automating at that price. Your actual cost still depends on how much work the agent does. Model announcement.

ServiceNow is bringing more of its workflows into ChatGPT. The October 28 partnership expansion includes plans for IT investigation, record updates, application development, and security work. The MCP plugin is available; several bigger additions are scheduled for late 2026 or 2027. That distinction matters when you’re deciding what you can use today. What’s live and what’s coming.

Oracle is making a similar bet. Its October 27 announcement describes Oracle AI Workspace, under development to bring Oracle applications and data into ChatGPT. My take: established software companies increasingly want customers to reach their products through an AI conversation. That’s worth watching if your business depends on owning the interface. The announcement.