🎧 Prefer to listen?

When OpenAI launched its Codex hardware a few weeks ago — a physical macro pad that triggers AI coding actions with a button press — it felt like a novelty. A cool gadget for people who already used Codex. But something bigger was happening underneath. OpenAI was in the middle of consolidating every AI tool it makes into a single product. And if you’ve been using ChatGPT for quick questions, you might not have noticed that the ground shifted under you.

Here’s the short version: Atlas, OpenAI’s dedicated web browser, is dead — less than nine months after launch. Codex, their coding agent, got absorbed into a new product called ChatGPT Work. The old ChatGPT experience you’re used to? It’s now called “ChatGPT Classic,” hidden behind a dropdown menu. And according to Bloomberg, 10 million people are now using OpenAI’s agent products, nearly doubling usage from earlier this month.

That’s not a feature update. That’s a complete product rewrite. And it changes what “using AI” means for people who don’t write code.

The consolidation nobody’s talking about

If you opened the ChatGPT desktop app this week, you probably noticed something different. There are now three “views”: Chat, Work, and Codex. Chat is the old conversational experience — type a question, get an answer. Work is the new thing: an agent that takes a goal and runs with it for hours. Codex still exists as a separate view for code-focused tasks, but it’s powered by the same engine as Work.

The confusing part is the naming. “ChatGPT Classic” is what they call the old experience. It’s tucked behind a “quick chat” button on desktop and a dropdown on mobile. OpenAI isn’t forcing anyone to switch, but the interface makes it clear which direction they’re pushing: toward agent-based workflows, not back-and-forth conversations.

This matters because most people still use ChatGPT the way they did in 2023 — type something, read the response, type something else. That’s like using a smartphone only for phone calls. The technology can do so much more, but the interface was the bottleneck. OpenAI just removed the bottleneck.

What ChatGPT Work actually does (for non-coders)

I’ve been testing ChatGPT Work since launch, and the difference from regular ChatGPT is not subtle. You give it a goal — “analyze my Q2 budget, find the three biggest spending categories, and draft a summary email to my team” — and it handles the entire chain. It connects to Google Drive to pull the spreadsheet, runs the analysis, writes the email, and waits for you to review before sending.

The key change: it doesn’t stop after a few minutes. Previous AI tools would time out or lose context after a few exchanges. ChatGPT Work can stay with a project for hours, maintaining context across hundreds of steps. That’s not an incremental improvement — it’s a different category of tool.

The integrations are the real story for non-technical users. ChatGPT Work connects to Slack, Microsoft Teams, Google Drive, and SharePoint through plugins you authorize once. On desktop, it can read and modify local files and use a built-in browser for web research. There’s also a Chrome extension that lets it perform web-based tasks without leaving your browser.

If you’ve been following the AI agent space, this is OpenAI’s answer to Anthropic’s Claude Cowork. Both tools let you delegate entire workflows instead of individual questions. But OpenAI’s version plugs directly into workplace tools most people already have open all day — Slack, Teams, Google Drive — without requiring you to set up a separate environment.

The cost reality check

Here’s where it gets real. ChatGPT Work uses GPT-5.6, which runs in three performance tiers. The highest tier — the one you’ll want for complex workflows — costs $5 per million input tokens and $30 per million output tokens. A single research task that reads a document, analyzes it, and produces a summary can easily consume 100K+ tokens.

OpenAI’s subscription plans go up to $100 a month with built-in usage limits through a credit system. I tested three real workflows to see what they actually burn:

  • Competitor analysis: “Find five competitors to [product], compare their pricing pages, and summarize in a table.” Touched 12 web pages, produced a 2-page report. Roughly 8% of a Plus plan’s monthly credits.
  • Meeting prep: “Read these 3 documents in my Google Drive, pull out key metrics, and draft talking points.” Three files, about 20 pages. Roughly 5% of monthly credits.
  • Content calendar: “Research trending topics in [niche], draft 4 blog post outlines with SEO keywords, and save to Drive.” Ran for 40 minutes, used the browser extensively. Roughly 15% of monthly credits.

The takeaway: ChatGPT Work is powerful, but it’s not free. If you’re on a Plus plan ($20/month), you can run maybe 10-15 complex workflows before hitting your limit. Enterprise and Edu subscribers can set spend limits at the group or individual level, which is important if you’re managing a team.

Scheduled Tasks: the feature that changes your day

The most underrated feature in ChatGPT Work is Scheduled Tasks. Think of it as AI-powered cron jobs — you can set up recurring tasks that run on a schedule or when a monitored event occurs. These tasks keep going when you’re away from your desk and can be monitored from your phone.

For non-technical users, this is the feature that moves ChatGPT from “useful tool” to “actual coworker.” Instead of remembering to ask ChatGPT to do something every Monday morning, you set it up once and it runs automatically. The agent checks your calendar, pulls the data, drafts the report, and sends you a notification when it’s done.

What the Atlas death tells you

The fact that OpenAI killed Atlas — its dedicated web browser — in less than nine months tells you everything about their strategy. The browser was always a stepping stone. The real goal was an AI that works inside whatever you’re already using, not a separate app you have to open and manage.

This is the same pattern we’ve seen in every technology that went mainstream: the interface disappears into the background. You don’t “use electricity” — you flip a switch. You don’t “use the internet” — you open an app. OpenAI is betting that you shouldn’t have to “use AI” either. It should just be there, inside your existing tools, doing the work you delegate to it.

For non-coders, this is the moment to stop thinking of ChatGPT as a chatbot and start thinking of it as a workflow engine. The consolidation isn’t confusing — it’s OpenAI simplifying its product line so you don’t have to choose between five different tools. There’s one entry point, and it does everything.

What to do right now

If you’re on a ChatGPT Plus or Pro plan, open the Work view and try something small. Don’t start with a complex multi-hour project. Start with something you do every week — a report, a summary, a comparison — and see how the agent handles it.

If you’re on a free plan, you’ll need to wait for broader rollout. But use this time to think about which parts of your workflow could be delegated. The tools are getting better every month, and the people who learn to use them now will have a massive advantage when they become standard.

The 10 million user milestone isn’t just a number. It’s a signal that agent-based AI has crossed from “interesting experiment” to “how people actually work.” OpenAI’s product consolidation makes that transition easier — fewer apps, fewer choices, one place to do everything. Whether that’s exciting or terrifying depends on how you feel about an AI that can work for hours without checking in.

The bottom line

OpenAI’s product shake-up — Atlas sunset, Codex merger, ChatGPT Work launch — isn’t about simplifying their product line. It’s about redefining what “using AI” means. The chatbot era is over. The agent era is here. And you don’t need to write code to use it.

If you want to see how other AI agents compare, check out our breakdown of AI agents for solo builders or start here to find the right tool for your workflow.