While you were offline this weekend, three things happened that change how your week starts.

Anthropic accidentally leaked their next model. ChatGPT learned how to write inside your apps. And OpenAI quietly finished training the thing that comes after GPT-5.

Most people will catch up to this by Friday. You're reading it now.

Here's what you need to know before your week starts.

On My Radar: 3 things I'm watching this week

Every few months, an AI company drops a new model and calls it "better." Usually it's 5-10% faster or a little sharper on benchmarks. You barely notice.

This one is different.

Over the weekend, Anthropic accidentally published details about a model called Mythos — codenamed Capybara — through a website error. Before they pulled it down, people saw the numbers.

The word used by researchers who tested it: "step change." Not incremental. Not gradual. A jump. Coding performance went up dramatically. Reasoning scores went way past anything Claude currently does. Even cybersecurity tasks — where AI usually struggles — showed huge gains.

Here's why this matters to you: Anthropic didn't announce this. They weren't ready. Which means they're building something they consider a full tier above Opus — their best current model. And it's far enough along to leak.

I've been covering AI for two years. Accidental leaks like this usually mean a launch is weeks away, not months.

What to do with this: If you use Claude for work — keep doing what you're doing. When Mythos drops, your same workflows get dramatically better overnight. Bookmark claude.ai and check back weekly. → claude.ai

The Signal: The AI you're using right now is about to feel slow by comparison. That's not a problem — it's an upgrade you didn't have to pay for. → Which makes Story 2 even more useful.

If you read about AI all week but you still haven’t figured out, how to actually use AI to automate tasks or achieve something.

You can describe your role and what you do to AI Insiders Assistant and it will help you automate upto 90% of your tasks w/ ai.

Until last week, ChatGPT lived in its own box. You asked it something, copied the answer, then pasted it into whatever app you actually use.

That middle step just disappeared.

OpenAI updated their ChatGPT integrations for Notion, Dropbox, Box, and Linear. The key change: ChatGPT can now write directly into these tools. Not just read your files — create new ones, edit existing ones, and update your project boards without you switching tabs.

This is the part that surprised me: it also shares your location now (if you allow it), which means it can give answers based on where you are — local events, nearby services, time-sensitive recommendations.

The people who connect their apps this week will feel like they hired a part-time assistant. Everyone else will keep copy-pasting.

What to do with this: Open ChatGPT → Settings → Connected Apps → turn on the ones you use daily. Start with Notion or Dropbox. Takes 3 minutes. → chat.openai.com

The Signal: AI tools are racing to be the one that lives inside your workflow — not beside it. The first one you connect wins your habit.

Here's a number that puts this week in context: $25 billion.

That's OpenAI's annualized revenue right now. A year ago it was $5 billion. Five times growth in 12 months. They're also talking about an IPO — possibly late 2026.

But the real story is quieter. On March 25, OpenAI finished pretraining their next model, codenamed "Spud." Pretraining is the heavy lifting — the part where the model learns everything. What comes next is fine-tuning and safety testing before release.

Nobody knows exactly what Spud will be. But finished pretraining means the countdown to launch has started.

Between Anthropic's leaked Mythos and OpenAI's freshly trained Spud — the next 60 days could bring two major model jumps back to back.

What to do with this: No action needed yet. Just know: whatever AI can do for you today is the worst it will ever be. Build your habits now so the upgrades land on something real. → aiinsidershq.com

The Signal: Two of the biggest AI companies just signaled major upgrades within weeks of each other. The gap between "using AI" and "not using AI" is about to get wider. Fast.

Three things I'm watching that could become bigger stories:

→ Anthropic's next move on Mythos. The leak happened. Do they accelerate the announcement or go quiet? If pricing shows up this week, it's coming sooner than anyone expected.

→ OpenAI's impact surveys start today (March 31). They're formally studying how their models affect the economy. Whatever they find shapes the regulation conversation for the rest of 2026.

→ Google Gemini 3 Deep Think just went live for Ultra subscribers. Early reports say it's strong on complex reasoning. If it holds up — Google's back in the conversation this week.

Move 1: Claude or ChatGPT → Weekly planning assist

Move 2: ChatGPT → Connect your first app