Daily AI Newsletter

The Tech Roundup

March 22, 2026

Hey {{{FIRST_NAME|there}}},

AI stopped asking for your permission this week — and the biggest names in tech just made that kind of power free.

🔍 The Big Story

AI Can Now Control Your Computer — And It Just Got Free

For the past few years, AI assistants like ChatGPT have been very good at telling you what to do: "Here's how you'd write that email." "Here's a step-by-step plan." But you still had to do the actual clicking, copying, and typing yourself.

That just changed. OpenAI's latest model, GPT-5.4, now includes native "computer use" — meaning ChatGPT Plus users can hand over a task and watch the AI navigate their computer, fill in forms, open apps, and get it done. You describe the job; it does the work. (Think of it as giving the AI a mouse and keyboard, not just a notepad.)

At the same time, Google released Gemini 3.1 Pro with deep integration into Gmail, Google Docs, and Sheets. If you live in Google's ecosystem, Gemini can now draft email replies in your voice, summarize long document threads, and make edits directly inside your files — no copy-pasting back and forth.

And Anthropic (the company behind the Claude AI assistant) just made their most capable model, Claude Sonnet 4.6, available on the free tier — with a context window of 1 million tokens. A "context window" is basically how much information the AI can hold in mind at once. One million tokens is enough to feed it an entire book, a year of emails, or a massive report and ask intelligent questions about all of it. Previously, that kind of capability cost money.

These three updates aren't random coincidences — they're part of an intensifying race to make the most powerful AI tools both accessible and genuinely useful for everyday tasks.

What can I do with this?

  • If you use Gmail, Docs, or Sheets, open Google Workspace and look for the Gemini button — try asking it to summarize your last 20 emails or draft a reply to a thread.
  • If you want to explore Claude for free, head to claude.ai — the free tier now includes Sonnet 4.6. Try uploading a long document and asking it to pull out the key points.
  • If you're a ChatGPT Plus subscriber, check the new computer-use features for automating repetitive tasks on your desktop or browser.

📋 In Brief

iPhone's Siri Is About to Get a Serious Upgrade — Powered by Google

If you've ever muttered in frustration at Siri for failing to understand a simple request, this is for you. Apple is set to release a completely redesigned Siri with iOS 26.4 this month — one powered under the hood by Google's Gemini AI model.

The new Siri will be able to understand what's happening across your apps, read the context of your conversations, and take real action — like booking a dinner from a restaurant someone mentioned in your texts, or setting reminders from an email without you having to spell it out. Crucially, Apple says all of this runs through their Private Cloud Compute system, a setup designed to process your data without exposing it to Google or Apple's servers.

The Apple-Google partnership is surprising given the two companies' long rivalry. But it signals something important: even Apple, which spent years building its own AI, decided the gap between its homegrown tools and the best external models was too wide to ignore.

What can I do with this?

  • When iOS 26.4 is available, update your iPhone and give the new Siri a real test — ask it to do multi-step tasks ("set a reminder from this email" or "book something based on this text thread").
  • Keep an eye on Apple's privacy documentation to understand what does and doesn't leave your device.

The White House Just Made Its First Big Move on AI Rules

The Trump administration released a seven-part national AI legislative framework this week — the most significant step the U.S. federal government has taken to shape how AI gets regulated. The proposal calls on Congress to create unified federal rules for AI, and importantly, to override state-level AI laws that the administration considers too restrictive.

That last part is the most controversial. More than 50 Republicans signed a letter warning that sweeping away state regulations could shield AI companies from accountability when their products cause harm. States like California and New York have been the most active in trying to set AI safety rules, and those laws could be wiped out if federal preemption passes.

The framework also includes provisions on child safety, limits on how much developers can be sued, and keeping electricity costs low for AI data centers.

What can I do with this?

  • Follow your state legislators to see how they're responding — this fight over who regulates AI (federal vs. states) will directly affect what protections you have as AI gets used in healthcare, hiring, housing, and more.
  • The NBC News report has a clear breakdown of all seven parts.

⚡ Quick Hits

The hidden human cost of your AI tools: A new UN report details the difficult working conditions of the largely invisible workforce that trains AI — content moderators and data labelers, many in the Global South, who review traumatic material, face intense monitoring, and earn low wages. The report calls for AI governance built around human rights. (United Nations)

Morgan Stanley: most people aren't ready for what's coming: The investment bank issued a stark warning that a major AI leap is imminent in early-to-mid 2026. AI models already perform at human-expert level on many economically valuable tasks, and the bank predicts faster workplace disruption in the next 12–18 months than in the previous five years combined. (Fortune)

Atlassian cuts 1,600 jobs for AI pivot: The company behind Jira, Confluence, and Trello is laying off 10% of its workforce to redirect resources toward AI features and enterprise sales. The company said AI "changes the mix of skills needed" — a phrase you'll be hearing a lot more this year. (Crescendo AI)

💡 One Thing to Try This Week

Give Claude's free tier a real workout with a long document.

Anthropic just made Claude Sonnet 4.6 free, with the ability to process up to 1 million tokens at once — that's roughly 750,000 words. Here's a practical way to use that right now:

  1. Go to claude.ai and sign in (or create a free account).
  2. Find a long document you've been meaning to read — a research report, a lengthy contract, a book chapter, or even a year of company meeting notes.
  3. Upload it or paste it in.
  4. Ask: "What are the 5 most important points in this document?" or "Are there any decisions or commitments made here I should know about?"

What used to cost money — or require you to actually read 100+ pages — now takes about 30 seconds and is free. It's one of those moments where AI genuinely saves you real time today, not in some hypothetical future.

You're reading The Tech Roundup — AI news for real people.
Forward to a friend  |  Unsubscribe