Hello AI Enthusiast,

This week OpenAI unveiled ChatGPT Work and, in the same breath, said goodbye to Atlas, the browser it launched less than a year ago. Meanwhile, Anthropic gave Claude users a new way to look back at how they've been using the tool. From ambitious pivots to a bit of self reflection, let's get into it.

The Big Picture 🔊

OpenAI launched ChatGPT Work

OpenAI launched ChatGPT Work, an agent inside ChatGPT powered by GPT-5.6 that gathers context from your apps and files and hands back finished documents, spreadsheets, presentations, or websites instead of a chat reply. It can work independently on a task for hours, is bundled with Codex into one desktop app, and rolls out across web, mobile, and desktop as OpenAI's answer to Claude Cowork.

Helin Yontar Helin Yontar CPO and Polyglot

Our founder Gianluca got early access and put it through a real test, editing a client proposal from his phone on the way to dinner, then turning an overnight holiday research task into a full interactive website. His verdict: ChatGPT Work isn't a big leap in raw intelligence over Claude Cowork, but it strips away enough friction that a much bigger crowd will actually use it.

That's the real story here. Anthropic proved agentic work could exist, OpenAI is about to hand a simpler version of it to over a billion people. Gianluca's bigger worry is the widening gap between companies that already know how to use AI well and those that haven't bought a single license yet. Agentic work also burns through tokens fast, so budgeting and teaching people to spend that budget wisely is quickly becoming its own skill.

His full breakdown, with screenshots of the AI-built travel site, is linked in the From Our Channels section below.

OpenAI Retires Atlas, Moves Browser AI Into ChatGPT

OpenAI is shutting down its Atlas browser on August 9, 2026, less than a year after launch. Browser-based agentic features like multi-tab navigation, downloads, and account logins are moving into ChatGPT and Codex instead. Users have about 30 days to export bookmarks and history before Atlas stops working for good.

Gioele Mottarlini Gioele Mottarlini COO and Image Addict

There isn't much to unpack here, it reads more like a support article than real news. But zoom out and it tells a story. Atlas was one of OpenAI's bets that AI could go head to head with search and browsing directly. That bet didn't stick.

Now the features are folding into ChatGPT, and OpenAI's energy is clearly shifting toward work and enterprise instead.

Anthropic Launches Reflect

Anthropic launched Reflect, a beta dashboard in Claude's Settings that summarizes your usage: key topics, active hours, and the kinds of tasks you hand off. It periodically asks questions like "What's one thing you want to keep doing yourself, even if Claude could do it faster?" and lets you set quiet hours. Reflect requires Memory to be turned on and is rolling out first on web and desktop.

Ross Horn Ross Horn Sales Guru

This feels like smart branding as much as a product feature. Where OpenAI's pitch has leaned toward "you don't need much skill, AI handles it," Anthropic is doing the opposite: telling people they still matter, and that how you use AI shapes what you get out of it. We especially liked that question about what you'd want to keep doing yourself. That's close to what we've been asking people in our bootcamps for years, so it's nice to see one of the leading AI labs land on a similar idea.

There's practical value too. Most people never touch features like Projects or Skills once they're comfortable with the basic chat, and a dashboard that notices your patterns and nudges you toward better tools could actually change that. It also taps into something people already like, the same instinct behind screen time or fitness trackers.

That said, requiring Memory to be switched on is a real adoption hurdle, and it's fair to note this also doubles as a way to keep people more engaged with the product, not just more reflective about it.

Bits and Bobs 🗞️

  • Anthropic is engaging the public to gather and address key questions and concerns about AI’s impact on jobs, society, and science to ensure the technology benefits everyone.

  • OpenAI emphasizes that effective AI investment depends not just on lower token costs but on matching model capabilities, governance, and workflow integration to maximize value and control risks in enterprise use.

  • Claude Code uses adjustable effort levels and model sizes to balance thoroughness, speed, and cost when generating AI code and debugging solutions.

  • Apple accuses OpenAI of stealing trade secrets and confidential information related to unreleased technologies and products, leading to a legal dispute amid their partnership.

  • Meta's new Muse Spark 1.1 is a powerful multimodal AI model that excels in coding, tool use, and managing complex tasks with improved speed, context handling, and safety features.

  • SpaceXAI's new Grok 4.5 AI model excels at coding, engineering, and knowledge work with faster, more efficient reasoning and is now available for free trial.

From Our Founder’s Channels 🤳

In his latest Substack piece, Gianluca Mauro shares his early hands-on experience with ChatGPT Work. Read it here.

ChatGPT Work lets your team take on more ambitious tasks, if they know how to use it. Good news: we know how to teach them. We have trained hundreds of non-technical people on how to perform ambitious agentic work with Codex, and ChatGPT Work makes it far easier.

That's a wrap on our newsletter! Before you go, here’s a quick recap of our offerings:

Catch you next week! 👋

Keep Reading