Claude crashes Microsoft's Office party; Google turns Gemini into your personal knowledge base; Anthropic launches a platform that lets any business deploy AI agents in days; and Anthropic's unreleased model found bugs that humans missed for 27 years - and got the Fed and the US Treasury into an emergency room together. It's been a week.

Hi! Welcome to the 44th edition of the TomorrowToday newsletter.

We’re here to decode the AI chaos so you don't have to. Think of us as your friendly neighbourhood tech translators - we cut through the chaos, translate the jargon, and spotlight new AI tools that matter for founders, builders, and curious minds.

Buckle up, because the future's moving fast and we're here to make sure you don't get left behind! ⚡

If you enjoyed today’s newsletter, please forward it to a friend & subscribe by following this link.

_{~9 mins read}

/Claude /Productivity /Word

If you've been using Microsoft's Copilot inside Word and wondering why it feels like a half-finished renovation, you're not alone. Microsoft has been promising AI-powered document editing for years, but the reality has often been clunky and frustrating to work with. Now Anthropic - the people behind Claude - have launched Claude for Word, and it's aimed directly at where Copilot falls short.

Think of it like this: instead of opening a separate browser tab and copy-pasting your document back and forth, Claude now lives inside your Word document. You can highlight a paragraph and say "make this more concise" and Claude edits just that bit, keeping all your formatting, numbering, and styles intact. You can ask it to summarise a 40-page contract in plain English. You can type "work through all five reviewer comments as tracked changes" and Claude will go through each one and propose edits - which you can then accept or reject, exactly like you would with a human colleague's revisions. It can even answer questions about your document content, with clickable citations that jump you straight to the relevant section.

Right now it's aimed at lawyers and finance professionals - people buried in contracts and dense documents all day. But honestly, anyone who works with long Word files will find this useful. The one catch? It's currently only available on Team and Enterprise plans, so personal subscribers will need to hold on a little longer.

/Google /Gemini /Productivity /NotebookLM

If you've ever started a research project in Gemini, lost your chats three days later in a sea of other conversations, and started from scratch - Google heard you.

This week, Google began rolling out Notebooks in Gemini - a way to organise your chats, uploaded files, and documents into dedicated project spaces. Think of it as a personal knowledge base that lives inside Gemini. You create a Notebook, drop in all your relevant files and previous chats, give Gemini some custom instructions, and from that point on it answers your questions with that specific context in mind. No more repeating yourself. No more starting from zero.

The clever part is the sync with NotebookLM - Google's AI research tool. Add a source to your Notebook in Gemini and it automatically appears in NotebookLM too. This means you can use Gemini for quick questions and chat, then switch over to NotebookLM to generate a podcast-style audio summary or a visual infographic from the exact same material. Two tools that actually talk to each other - which, in the tech world, is rarer than you'd think.

It's rolling out to Google AI Ultra, Pro, and Plus subscribers on the web first, with free users and mobile getting access in the coming weeks.

/Claude /Agents /Enterprise /Productivity

Before we get into the news, a quick explainer - because the word "agent" gets thrown around constantly and most people still aren't sure what it actually means.

Here's a simple way to think about it: a chatbot is like a very smart friend you can text questions to. An AI agent is like hiring a personal assistant who never sleeps, doesn't need coffee, and can actually do things on your behalf - not just talk about them. Instead of asking "how do I submit my expenses?", you tell the agent to submit your expenses - and it figures out the steps, does them, and checks back only when it hits something it's unsure about. Think of it like a specialist who is excellent at this one particular job, and does it relentlessly, around the clock.

Until now, building one of these agents has been a monster of a technical project. Your developers needed to set up secure environments, manage long-running sessions, handle errors, control permissions, and basically build a whole infrastructure before writing any of the actual logic. That could take months.

Claude Managed Agents - launched this week in public beta - removes all of that. You describe what you want your agent to do, what tools it can use, and what guardrails it should operate within. Anthropic's platform handles the rest: the secure sandbox it runs in, the session management, the error recovery, the scaling. Companies like Notion, Asana, and Sentry are already live on it. Rakuten stood up enterprise agents across five departments - product, sales, marketing, finance, and HR - in under a week per team. Sentry paired it with their existing debugging tool and went from flagged bug to reviewed code fix automatically, shipping that integration in weeks instead of the months they'd projected.

Anthropic claims the platform cuts development time by up to 10x. Whether or not that figure holds in every situation, the direction is clear: deploying an AI agent is moving from a six-month engineering project to something a small team can do in days.

In this section, we aim to spotlight an incredible AI tool or use case and guide you on how you can try it.

Claude now lives inside Microsoft Office - and it's genuinely impressive once it's up and running. Here's how to install it, and two things to try immediately.

⏱ Time to set up: about 10 minutes

Installing Claude in Word:

Installing Claude in Excel:

Try this in Word - instant document summariser:

Open any long document, activate the Claude panel, and paste this in:

Claude reads the whole thing and gives you a structured summary, with clickable references to the exact sections it's drawing from.

Try this in Excel - no-fuss data analysis:

Open a spreadsheet with some data - sales figures, a budget, survey results, anything with numbers - activate the Claude panel, and try this:

Claude reads your spreadsheet, writes the formulas, and builds the summary table. No pivot table wrestling. No hours lost to formatting. Just results. ⚡

In this section, we're spotlighting real businesses using AI to solve actual problems.

Right, let's talk about something that's been making headlines in Washington this week and that most people outside the tech world haven't fully grasped yet.

Anthropic has a new AI model called Claude Mythos. They've built it. They've tested it. And they're not releasing it to the general public - at least not yet. That alone should get your attention.

The reason? Mythos is, by a significant margin, the most capable AI system ever built for finding hidden flaws in software. And Anthropic is worried about what happens when that capability ends up in the wrong hands.

So what did it actually find?

Finding software bugs has traditionally required years of expertise and thousands of hours of painstaking work. Mythos found thousands of them - in some of the most widely-used software in the world - in a matter of weeks. Here are three examples that should make you sit up:

All three have since been patched. But the point isn't the specific bugs. The point is the speed and independence with which Mythos found them - problems that human experts missed for decades.

Serious enough to call an emergency meeting.

Federal Reserve Chair Jerome Powell and Treasury Secretary Scott Bessent called the CEOs of America's biggest banks - Citigroup, Bank of America, Morgan Stanley, Wells Fargo, and Goldman Sachs - into an urgent meeting specifically to discuss the cyber risks that a model like Mythos represents. Not about interest rates. Not about tariffs. About an AI model.

Now, Anthropic has a reputation for being cautious - sometimes to the point where critics call it scaremongering. They've been accused of raising alarms about AI risks while simultaneously building the very models they warn about. That tension is real, and it's fair to hold onto healthy scepticism.

But this time, it feels different.

We are, quietly, crossing a threshold. The gap between what the most advanced AI models can actually do, and what the general public believes they can do, has never been wider. Mythos isn't a chatbot that helps you write emails. It's an autonomous system that can find hidden vulnerabilities that humans - and millions of automated tests - missed for decades. When that kind of capability becomes widely available (and eventually, it will), the cybersecurity landscape changes permanently.

Anthropic's response - under the banner of Project Glasswing - is to use Mythos defensively first. They've brought in AWS, Apple, Google, Microsoft, NVIDIA, JPMorgan, and others to scan and patch critical infrastructure before the capability spreads. They've committed up to $100 million (roughly R1.8 billion) in usage credits to support the effort.

It's a smart move. Whether it'll be enough is another question entirely.

AI is full of jargon, and we’re here to decode it. Each week, we’ll give you a plain-English definition of a buzzy term you’ve probably seen (but never fully understood).

Thanks for reading TomorrowToday! We’d love to hear from you:

➡️ What would you like us to cover next?
➡️ Have a tool or topic we should feature?

We’re building this with (and for) you. 🚀
See you next Tuesday 👋