Window Into the Future 🔮

So here's what happened at Anthropic in the last seven days alone.

• Claude got computer use which means it can now open your apps, navigate your browser, fill in spreadsheets, and operate your desktop like a colleague borrowing your laptop.

• There's a new tool called Dispatch that lets you assign Claude tasks from your phone and come back to finished work.

• Fortune broke a story about Mythos, a next-generation model that leaked accidentally and was described internally as a "step change" in capability.

• And reports surfaced that the company is in early talks about going public, potentially as soon as October.

Any one of those would've been the biggest AI story of the week but I don't want to just walk you through the news. I want to talk about what the pace itself is telling us because I think many people are drawing the wrong conclusion from it.

The obvious reaction to a calendar like that is I need to keep up. Learn all these features. Understand computer use and Dispatch and the new model and what it all means for my workflow. That reaction makes sense because it's how we've always thought about technology.

But the people I know who are genuinely good with AI are not keeping up with the releases. What they've done instead is go deep on one or two things. They have a Claude project that knows their entire context. They have a workflow they've iterated on dozens of times. They've built something specific to how they work, and they keep refining it.

There’s research backing this up. BCG found that only about 5% of companies are generating real value from AI at scale despite adoption rates hitting 88%. They call it the adoption puzzle: usage is up, impact isn't. Because most orgs treat AI like a software rollout where they give people access, run a training session, and move on when what actually works looks more like building a habit.

The calendar is impressive, but it's slightly misleading if you're a professional trying to figure out what to do with it. Because it implies that the person who knows about all 52 releases is winning. My honest read is the opposite. The people pulling ahead right now are the ones who picked one use case three months ago and just kept going back.

I've been living at Network School for the last month with 400+ founders, creatives, freelancers all in one place and it honestly feels like eavesdropping on the future. One person here has seven paid Claude accounts. SEVEN. Most have more than one. I was surprised at first, but I think for most of us our imagination just hasn't caught up to what's possible with that much usage. One account versus seven is a completely different capability. And it made something click for me: the people who are getting disproportionate value aren't dabbling across every AI tool. They've gone absurdly deep on one.

And this is worth being honest about too. Wharton researchers recently coined the term "cognitive surrender" for what happens when people stop thinking critically about AI output. With AI, the answers are right often enough that you stop checking, and your judgement declines. But the part I find genuinely hopeful from the research is that when an AI tool was designed to coach rather than just answer, the negative effects disappeared entirely. How you use the tool matters more than whether you use it.

So the people who go deep develop the judgement to know when Claude is helping and when it's wrong. The people who stay shallow never build that muscle. And in a world where your AI tool updates every few days, that muscle is the only thing that compounds alongside the technology.

Every one of those releases is making the tool better at the thing those people are already doing. Computer use in Claude means the workflow they've been building can now go further. Memory means their Claude project retains more context. The 1M context window means they can feed an entire quarter's worth of work. These aren't 52 things to learn but 52 upgrades to a practice you already have (if you have one).

And the Anthropic calendar isn't just a to-do list for power users. It's a window into what an AI-first company actually looks like when it's compounding. 80% of Anthropic employees use Claude daily. One employee alone hit $150,000 in Claude Code usage in a single month. The tool is building itself and shipping at a speed that most orgs won't reach for years.

You're watching what the future of work looks like.

I know this can sound like "if you're not using AI daily you're falling behind" and that framing is exhausting. It's not what I mean. There's a difference between trying to keep up with the news and building something real with the tools.

The first one will burn you out. The second one will genuinely change your work. And a lot of what's on that calendar (the Excel integration, the PowerPoint add-in, Cowork, computer use) is Anthropic meeting you where you already are. They're integrating Claude into the apps you already have open daily. If you've been waiting for a good time to start, the barrier just got meaningfully lower.

How to AI 🤖 

Every week, this section is your shortcut. Since this entire issue is about Claude, here are some practical ways to put it to work this week:

• How to effectively use Claude Cowork

• How to design faster with Claude Skills

• How to build a personal command center in Claude Code

• How to use Claude Cowork to build custom agents, automations, and Skills

Before you go ✌️

What's the one thing you keep coming back to Claude for? Is it the quality of the copy/ output, the actual workflow, or something else?

Hit reply. I read them all.

See you next Sunday!