• Everyday AI
  • Posts
  • Ep 819: ChatGPT Work: What’s New, Who It’s For and How to Use It

Ep 819: ChatGPT Work: What’s New, Who It’s For and How to Use It

The Trump administration launched a new AI cybersecurity hub, OpenAI is reportedly building a smart-home speaker, and New York paused new AI data centers. And more.

Author

Everyday AI
July 15, 2026

Outsmart The Future

Today in Everyday AI
8 minute read

🎙 Daily Podcast Episode: OpenAI's ChatGPT Work makes Codex easier to use without removing its core AI agent capabilities. Here's what businesses need to know. Give today’s show a watch/read/listen.

🕵️‍♂️ Fresh Finds: Claude is giving teachers free AI tools, Anthropic is backing tougher AI safety laws, and Elon Musk says X will open source its code. And more. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: The Trump administration launched a new AI cybersecurity hub, OpenAI is reportedly building a smart-home speaker, and New York paused new AI data centers. And more. Read on for Byte Sized News.

💪 Leverage AI: ChatGPT Work makes OpenAI's AI agents easier to use by hiding the complexity behind Codex. Here's how to put them to work across your business. Keep reading for that!

↩️ Don’t miss out: Miss our last newsletter? We covered: Grok reportedly sent entire Git repos to xAI servers, Meta is facing an AI-related layoffs lawsuit, and OpenAI's GPT-5.6 arrived on Amazon Bedrock. And more. Check it here!

Ep 819: ChatGPT Work: What’s New, Who It’s For and How to Use It

OpenAI's new ChatGPT Work is on fire. 🔥

The company said they've added 2 million users in the past two days alone.

Wanna know a secret about it though?

(It's kinda not new at all. It's just Codex in Chat clothing.)

Regardless, millions of new users are realizing for the first time what it means to have an agent that can work for them 24/7. And even if you're not a technical builder, it's pretty easy to set up a swarm of agents to help with any kind of work.

Also on the pod today:

• ChatGPT Work vs Codex explained 🤔 
• Desktop vs web differences 🌐 
• Remote mode from your phone 📱 

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 –  ClawTeams is An IM-native AI staff collaboration platform, Branda is On-brand ads generated in seconds, Vocalvia Turns any document into an editable podcast episode.

Claude for Teachers — Claude for Teachers is giving US K-12 educators free access to AI-powered lesson planning and classroom tools, all mapped to state standards.

Anthropic State by State — Anthropic is pushing for the toughest AI safety laws yet, backing strict audits and risk checks in several states.

George Lucas AI — George Lucas calls out Hollywood for letting focus groups and fans steer movies, saying real storytelling should come from passionate filmmakers.

Superhuman AI Emails — Superhuman’s new AI drafts emails that actually sound like you, not a bot.

Meta Muse Spark 1.1 Arena — Meta’s Muse Spark 1.1 just debuted in Agent Arena, landing at #17 and beating Gemini 3.1 Pro.

Codex Micro — Work Louder’s Codex Micro lets you monitor ChatGPT agents with physical RGB keys.

Sesame Calendar — Sesame is adding Calendar and Email connectors for its iOS app, letting AI handle more for you.

Spotify Controversial AI Feature — Spotify’s new conversational feature lets you shape your music, ask questions, and explore your listening history just by talking or typing

X Open Source — Elon Musk says X will open source its entire codebase after a security review, promising full transparency.

Gemini in Chrome UK — Gemini is now built into Chrome for UK users, letting you chat with an AI assistant to summarize pages or draft emails right in your browser.

Apple AI — Apple is eyeing tech that shrinks massive AI models to fit on your iPhone, promising faster, more private features.

ChatGPT Cycling — After losing his leg as a child, Vishal used ChatGPT to train, plan nutrition, and tweak his prosthetic. Now he's set to represent India at two major para cycling championships.

1. Trump Administration Launches GOLD EAGLE Cyber-Patching Hub 🦅

The Trump administration has announced GOLD EAGLE, a new federal-industry clearinghouse designed to speed up the detection, verification, and patching of cybersecurity flaws affecting critical infrastructure.

Created under a June 2 executive order, the effort brings Treasury, DHS and CISA, the Department of War, open-source partners, and major infrastructure companies into a shared response system. The central bet is that frontier AI can help defenders sort urgent risks faster while cutting repeated scanning work.

2. OpenAI Reportedly Targets Smart-Home Speaker With Humanlike AI Personality 🔊

According to Bloomberg, OpenAI is developing a mobile, screen-free smart speaker that could mark its first major move into consumer hardware ahead of a potential IPO.

The device is designed to manage smart-home tasks, media, messages, and ChatGPT features while learning its owner’s habits and drawing on personal data such as email. Its moving mechanical parts and conversational style are meant to make it feel less like a gadget and more like a household companion.

3. Apple Tests PrismML’s Tiny iPhone AI Models 📱

According to CNBC, Apple is evaluating PrismML’s technology, which the startup says can compress a 27-billion-parameter AI model from about 54 GB to under 4 GB so it can run directly on newer iPhones.

The news lands just as Apple opens an iOS 27 beta featuring its revamped Siri, underscoring its push to make the assistant faster and more private without sending every request to the cloud. PrismML claims its compressed models use far less memory, respond much faster, and consume less power, though they give up some factual accuracy.

4. New York Pauses New 50-MW AI Data Centers for Up to a Year ⚡

Gov. Kathy Hochul has signed an executive order temporarily blocking new large-scale data centers that use at least 50 megawatts of power, citing pressure on the grid, water resources, and household utility bills.

The move makes New York the first state to impose a statewide pause of this kind, giving officials time to set rules for local siting, cleaner power, and consumer protections.

5. Publishers and Authors Sue Google Over Gemini’s Book Training 📚

The publishers and authors also claim Google altered or removed copyright information to hide the alleged use of protected material. The filing lands as courts weigh whether AI training qualifies as fair use, with early California rulings favoring AI companies but leaving major questions unresolved.

6. IBM Plunges 23% After Z17 Mainframe Miss Raises AI Spending Doubts 💸

IBM shares tumbled more than 23% at Tuesday’s open after its second-quarter profit and revenue fell short of Wall Street expectations, putting the company on track for its worst day in decades.

CEO Arvind Krishna said weaker-than-expected sales tied to the Z17 mainframe and its transaction-processing software were the central problem, despite the system’s AI-focused pitch. IBM also said customers redirected late-June budgets toward servers, storage, and memory ahead of expected price increases, leaving less room for its software and infrastructure purchases.

One word kept nontechnical workers outside OpenAI’s desktop agent: code.

ChatGPT Work changed the label, simplified the view, and two million users showed up in 48 hours.

Pay attention.

Codex already had computer access, browser use, files, apps, automations, long-running goals, and parallel subagents. The breakthrough was making all of that feel like normal work.

The competitive gap is simple: one team babysits a chatbot while another directs several agents across real workflows.

That gap can get ugly fast.

On today’s Everyday AI, we broke down the packaging play, the desktop-versus-web decision, and the management system leaders need before agentic work spreads faster than their controls.

1. OpenAI made Codex feel usable ⚡

ChatGPT Work on desktop is largely Codex behind a friendlier front door. Work mode hides the code-heavy detail; Codex keeps the diffs, changed files, and technical visibility.

Same engine, better invitation.

OpenAI removed the developer-first framing and pointed users toward finished documents, presentations, analysis, dashboards, and other business deliverables.

That should make leaders rethink every AI rollout built around licenses, feature tours, and vague encouragement to experiment. Access means little when employees can’t connect the tool to the work already on their plate.

Nah.

Buying capability and hoping adoption follows ain’t a strategy.

Try This: Rebuild one underused AI rollout around three finished outputs, not a feature list.

Give finance a variance brief, sales a researched account plan, or operations a live tracker. If employees still need technical jargon to get started, simplify the doorway again.

2. Move ChatGPT Work to desktop 🚀

ChatGPT Work on the web gives casual users a familiar place for longer jobs. Desktop creates the bigger shift because it can use the browser, computer, local files, apps, connected tools, Remote, and scheduled tasks.

One instruction can analyze beehiiv performance, inspect months of context, build a mobile-friendly site, publish it, and attempt the final handoff. That collapses a scattered chain of knowledge work into one directed outcome.

Remote lets you start, steer, or check that work from your phone while the machine handles execution elsewhere.

Add queues, loops, long-running goals, and subagents, and one employee can supervise several workstreams instead of waiting for one thread to finish.

This changes the job.

Your people move from prompting an assistant to directing specialists, which only works when the workflow, context, and finish line are clear enough to avoid constant rescue.

Try This: Choose one weekly process that crosses at least three systems and ends in a shareable deliverable.

Map the approved sources, output format, review point, and final action. Once the workflow holds up, split research, production, and QA across subagents.

3. Manage agents before they multiply 🔥

More capable agents create a management problem fast.

They can run for hours or days, rewrite files, use your computer, queue follow-ups, and create their own specialist teams. Goals define the finish line.

Plan mode exposes the approach, steering corrects the work midstream, and AGENTS.md files preserve the rules across the project.

Permissions still matter because approval friction is real and unrestricted computer access is reckless.

The companies that scale this well will decide what agents may read, write, publish, send, and spend before dozens of autonomous tasks create an expensive, untraceable mess.

Yeah, governance sounds less exciting than a viral launch. It also decides whether ChatGPT Work creates dependable capacity or polished mistakes moving at machine speed.

Try This: Write a one-page operating brief for the first workflow: goal, approved folders, connected apps, forbidden actions, output standard, reviewer, and usage ceiling.

Turn those decisions into an AGENTS.md file and require review before external messages, publishing, or irreversible actions. Then let the agent run.

Reply

or to participate.