- Everyday AI
- Posts
- Ep 696: Hands on with Claude Cowork: Is This the Future AI Interface?
Ep 696: Hands on with Claude Cowork: Is This the Future AI Interface?
Claude Cowork: How it works, New leaks on Apple's AI, Campos, Google confirms Gemini will remain ad-free, OpenAI unveils its “Education for Countries” initiative and more
Everyday AI
January 21, 2026
👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI
Outsmart The Future
Today in Everyday AI
8 minute read
🎙 Daily Podcast Episode: Most AI still waits for instructions. Claude Cowork doesn’t. It works inside your machine, moves files, uses your browser, and handles the busywork humans usually do between tools. Give it a watch/read/listen.
🕵️‍♂️ Fresh Finds: Claude adds new health features, Adobe Premiere Pro rolls out AI video tools, Gates Foundation and OpenAI launch a $50M AI health program for African clinics and more Read on for Fresh Finds.
🗞 Byte Sized Daily AI News: ChatGPT launches an age-prediction safety model, Google confirms Gemini will remain ad-free, new plans leak on Apple’s upcoming AI chatbot and more. Read on for Byte Sized News.
💪 Leverage AI: Claude Cowork isn’t a chatbot upgrade — it’s an AI that actually uses your computer. Copy-paste work just became optional. Keep reading for that!
↩️ Don’t miss out: Did you miss our last newsletter? We Covered: Anthropic Teams Up with Teach For All on Global AI Education, ServiceNow Finalizes Multi-Year OpenAI Partnership, Microsoft and BMS Expand AI Lung Cancer Detection Across Hospitals and more Check it here!
Ep 696: Hands on with Claude Cowork: Is This the Future AI Interface?
Wait.... AI can do THAT now? 🤯
(Yes, it can.)
Two things that consumer AI has never been able to do very well unless you were a cracked dev: access your computer and control your browser. Welp.... that's changed with Claude Cowork.
Now, you can combine the best of Claude and Claude Code with the nitty gritty human glue work that we find ourselves doing more and more in the age of AI.
Also on the pod today:
• Claude Cowork vs Claude Code 🤔
• AI takes over local files 💻
• Non-devs automating complex tasks ⚡
It’ll be worth your 36 minutes:
Listen on our site:
Click to listen
Subscribe and listen on your favorite podcast platform
Listen on:
Here’s our favorite AI finds from across the web:
New AI Tool Spotlight – Mastra is a framework for building AI-powered applications and agents with a modern TypeScript stack, Fimo Instantly generate syour site with AI, then collaborates with your team to keep it forever in motion, Citable Tracks and improve how different AI’s mention your brand to your customers.
Claude Health — Health data joins Claude—private by design, but trust still on the line
On-Device Function Calling — Tiny 270M FunctionGemma runs offline. Fine-tune it, accuracy jumps ~58% → ~85%.
Premiere Pro AI — Premiere just got instant object masking. After Effects now sculpts 3D scenes and animates fonts.
Youtube Short AI — YouTube will let creators make Shorts using AI likenesses. How will it prevent fake or low-quality clips?
Audio to Video — LTX turns audio into lip‑synced, beat‑timed video. Try it for consistent voices.
Health Program for African Clinics — Gates Foundation and OpenAI launch $50M AI health program for 1,000 African clinics. Learn more
AI Creativity — Google.org gave Sundance $2M to train filmmakers in AI, with free courses and a fellowship.
1. DeepMind says no ads in Gemini, for now 🇽
Google DeepMind CEO Demis Hassabis told reporters in Davos that the company has no plans to put ads into its Gemini AI, a timely reassurance as tech leaders gather and AI monetization debates heat up.
The comment matters because it signals Google’s current stance on user experience and trust amid industry pressure to commercialize large language models. Hassabis’s remark also comes as broader AI competition and regulatory scrutiny intensify, shaping how major players might balance revenue and safety.
2. ChatGPT adds age-guessing to protect teens 🛡️
OpenAI has begun rolling out an age prediction model in ChatGPT to flag accounts likely belonging to people under 18 and automatically apply stricter content safeguards, with the goal of reducing teen exposure to sensitive material.
The system uses behavioral and account signals such as account age, activity patterns, and stated age, and defaults to a safer experience when information is incomplete. Users misclassified as under 18 can quickly restore full access by verifying their age with a selfie through Persona, and parents can further customize protections with parental controls.
3. OpenAI launches “Education for Countries” to fold AI into national school systems 🇬🇷
OpenAI today announced Education for Countries, a new initiative partnering with governments and universities to integrate GPT‑5.2, ChatGPT Edu, study mode, and other tools into national education systems to personalize learning and prepare students for an AI‑shifted workforce.
The program pairs nationwide tool deployments and teacher training with large‑scale research and certifications to measure impact on learning and teacher productivity, and its first cohort includes eight countries such as Estonia, Jordan, and the UAE.
4. OpenAI signals first physical device coming in late 2026 📲
OpenAI leadership told Davos attendees that it is "on track" to unveil a first physical device in the second half of 2026, with senior executive Chris Lehane saying more details will come later in the year.
The company offered no specifics on form or function, though previous reports described a pocketable, screenless AI companion designed to coexist with users rather than replace phones. OpenAI’s CFO framed 2026 as a year of practical adoption, noting dramatic compute and revenue growth that underpins such product moves.
5. Report: Apple to turn Siri into a full chatbot this year 🍎
Apple plans to replace the current Siri with a chat-style assistant called Campos, aiming for a June reveal at WWDC and a September release as part of iOS 27, iPadOS 27 and macOS 27, according to people familiar with the plan. The chatbot, built on a custom Google Gemini-based model and deeply integrated into Apple apps and on-screen controls, will handle voice or text, access personal data for tasks, and run high-end models likely hosted on Google TPUs.
Privacy and memory limits remain under discussion, and Apple is designing the system so its underlying models can be swapped if it wants to move away from Google in the future.
Anthropic quietly shipped the first AI interface that does what you currently do.
Not what you ask it to do. What YOU do.
The copying. The pasting. The downloading, uploading, and navigating between 15 browser tabs while your AI just sits there waiting for the next file transfer. All that human connector work between AI tools? Claude Cowork now handles it.
So on today's episode of putting AI to Work on Wednesday, we went hands-on with Anthropic’s latest release to show you why it matters far more than the demos suggest.
Because what Cowork is today is not what it represents for tomorrow.
1. You Were The Duct Tape 🚀
Here's what nobody talks about when using AI tools. Yes, you're orchestrating. Yes, you're commanding. But between those high-value moments, you're basically a human USB cable.
Copy this output. Paste it there. Download that file. Upload it somewhere else. Navigate to this website because your AI can't browse.
Tedious.
Claude Cowork eliminates that entire middle layer. It connects your local computer, your terminal, and a Chrome browser that it actually controls. This is the first mainstream AI interface combining all three access points that humans currently manage manually.
The result shifts your job description entirely. You become the person setting direction and checking outcomes while Cowork handles everything between.
Try This: Map your last complex AI task and count every manual transfer between systems. Every copy, paste, download, and browser navigation represents work Claude Cowork handles autonomously.
Most knowledge workers discover 40% of their AI workflow involves these connector tasks. That percentage reveals exactly how much time you could reclaim.
2. Built By AI In Ten Days 🔥
Entirely.
100% of the product came from engineers orchestrating AI agents rather humans than writing code themselves. The whole thing shipped in roughly a week and a half. That timeline matters because it demonstrates exactly what Cowork enables for everyone else.
Still buggy? Absolutely.
Updates rolled out over the weekend and more will follow. But the underlying architecture already works well enough to create software, reorganize files across hundreds of folders, and navigate authenticated websites using your actual browser sessions.
At the time of recording our show, Claude Cowork required the $100/mo Max plan, but they immediately rolled out access to $20/mo users right when we finished. (Yeah, that’s how it works sometimes.)
The desktop app is only on Mac for now but Anthropic hinted it may roll out for other platforms oon.
Try This: Identify one disposable tool you wish existed for a specific workflow. Something simple like we demoed like a Kanban board for today's tasks instead of a random sticky note on your desktop.
Instead of searching for software that almost fits, consider building exactly what you need in minutes. Custom disposable software is arriving faster than most realize.
3. Exploring Buggy Tools Creates Advantage ⚡
Guaranteed.
The question is whether you understand its capabilities when that happens or scramble to catch up while competitors already integrated it into their workflows.
Every major AI interface shift rewards early exploration. Not early perfection. Exploration.
Right now Claude Cowork can access anything on your local machine, run terminal commands without you touching command lines, and control Chrome using the Claude Code extension. It carries over your cloud memory and skills from regular Claude conversations. These building blocks enable workflow automation that previously required technical expertise.
The companies treating this as experimental curiosity will wonder how competitors suddenly move faster. The ones exploring while it's rough will already know exactly how to deploy it when polish arrives.
Try This: Schedule 30 minutes this week to explore one capability you've never attempted with AI tools. Maybe browser automation or mass file organization.
The specific task matters less than building muscle memory for this interface paradigm before it becomes mainstream.
Reply