March 25, 2026

The Tech Roundup

AI news for real people

AI just graduated from "thing you talk to" to "thing that does things for you" — and it happened all at once this week.

🔍 The Big Story

Your AI Can Now Use Your Computer While You're Away

Imagine sending a message to your assistant before you leave for lunch: "Book me a dentist appointment, update the project spreadsheet, and order more coffee pods." When you get back — it's done. All of it. That's the future Anthropic unveiled this week.

Claude, Anthropic's AI, can now act as a full computer-use agent for everyday people. You send it a task from your phone, and it opens apps on your desktop, navigates your browser, fills in forms, and completes the job on your behalf — no coding required, no technical setup. This is a major leap from AI as a chatbot (something you talk to) toward AI as a personal assistant that actually does things for you.

The practical shift here is huge. Until now, automating computer tasks meant learning scripting tools or hiring someone to set it up. Anthropic is betting that the average person — not just developers — is ready to delegate their to-do list to an AI that can operate their computer like a remote employee.

This is early technology and it'll make mistakes. But the direction is clear: AI agents that handle your repetitive digital work while you focus on what actually needs your attention.

What can I do with this?

  • Try delegating a tedious task you do weekly — filing documents, copying data between spreadsheets, booking a recurring appointment — and see how Claude handles it
  • Start simple: one-step tasks with clear instructions work best right now
  • Visit claude.ai on desktop — computer use is rolling out gradually, so look for the agent or automation options in the latest version (you may need to opt in or join a waitlist)

📋 In Brief

The White House Wants One Set of AI Rules for All 50 States

The Trump administration released a National AI Legislative Framework this week, asking Congress to create a single federal standard for how AI companies can operate. The biggest impact: it would override state laws — meaning stricter rules that places like California have passed could be wiped out. The framework also proposes protecting AI developers from excessive lawsuits and adding child safety requirements on AI platforms.

This matters because right now, AI rules are a patchwork — different states have different protections around AI-driven decisions in hiring, lending, and healthcare. A federal standard simplifies things for companies but may weaken protections in states that went further.

What can I do with this?

  • If you care about AI decisions that affect your job or finances, look up what AI protections your state currently has — they may not last
  • Follow this story: Congress still has to act on it, and the debate will shape AI rules for years

OpenAI Upgrades Free ChatGPT — and Adds Live Math Modules

Starting this week, free ChatGPT users get access to a new "Thinking" mode powered by GPT-5.4 mini. This is a more capable reasoning model that walks through problems step-by-step before answering — useful for anything that requires real analysis, not just a quick response. At the same time, legacy "deep research" mode is being retired on March 26.

Bonus for students: ChatGPT now includes interactive math and science modules. Ask about a formula and you'll get a live visual tool where you can plug in numbers and see results change in real time. It's like a calculator that also explains the math.

What can I do with this?

  • Free users: tap the + menu in ChatGPT and look for the "Thinking" toggle — you now have access to a smarter reasoning mode at no cost
  • Students or anyone learning a new skill: try asking ChatGPT to explain a concept and use the new interactive modules to play with it hands-on

⚡ Quick Hits

Siri is getting a brain transplant: Apple confirmed the new Siri — powered by Google's Gemini AI model with Apple's privacy protections baked in — arrives with iOS 26.4 in the coming weeks. For hundreds of millions of iPhone users, this could finally make Siri worth talking to. (PYMNTS)

AI has automated 12% of job tasks — but created new ones too: A McKinsey Global Institute report found AI displaced 12% of tasks across the economy over two years, but also sparked 8% growth in entirely new job categories like prompt engineers, AI trainers, and model auditors. It's not "AI takes all jobs" — but roles built on repetitive, rules-based work are most at risk. (Crescendo AI)

Perplexity wants to live on your desk permanently: The search-AI company announced "Personal Computer," a dedicated AI agent designed to run continuously on a local device (think Mac Mini). It accesses all your files and apps, runs privately without the cloud, and works autonomously throughout your day. It's a new category — less assistant, more always-on AI coworker. (TLDL)

💡 One Thing to Try This Week

Turn on ChatGPT's "Thinking" mode — it's free now.

If you use the free version of ChatGPT, you've been working with a capable but speed-focused model. This week, OpenAI opened up GPT-5.4 mini in "Thinking" mode to free users — and it's noticeably better for anything that requires reasoning: planning, comparing options, working through a problem.

Here's how to try it in under two minutes:

1 Open chatgpt.com or the ChatGPT app
2 Tap or click the + button near the message bar (or tap the model name shown at the top of the chat)
3 Select Thinking from the list
4 Ask it something that requires real analysis — "Help me decide between these two job offers" or "What should I know before refinancing my mortgage?"

You'll see it work through the problem before answering — and the answers are noticeably more thorough. No subscription required.

You're reading The Tech Roundup — AI news for real people.

Forward to a friend   |   Unsubscribe