What OpenAI Added — and What It Quietly Took Away
OpenAI’s spring 2025 product push had a clean narrative: the company was moving fast, shipping powerful new tools, and pushing ChatGPT into serious enterprise territory. Codex, its cloud-based agentic coding tool, landed as a flagship release. ChatGPT Work rolled out as a dedicated workspace for professional teams. The press releases wrote themselves.
What didn’t make the announcements was the other side of the ledger.
While OpenAI was framing these launches as pure additions, the ChatGPT Mac desktop app was getting quietly hollowed out. Features that everyday users relied on were removed or degraded — not deprecated with fanfare, not flagged in a changelog users would actually see, just gone. The desktop experience that had built a loyal base of regular users took a step backward at the exact moment the enterprise pitch stepped forward.
This is a pattern the tech industry has refined over decades. Companies announce additions loudly and walk back subtractions in silence, burying regressions beneath the noise of launch events. OpenAI executed it cleanly here. The headline was Codex and ChatGPT Work. The footnote was a worse desktop app for the millions of non-developers who use ChatGPT to write, think, and get through their day.
The result is a split product. For developers and enterprise teams, OpenAI is genuinely building new capability. For the everyday ChatGPT user — the person who made the AI chatbot a household name, who drove it to 100 million users faster than any consumer app in history — the desktop experience now pushes them back to the browser. That’s not a neutral outcome. It signals where OpenAI’s attention and resources are actually flowing.
The ChatGPT desktop app was supposed to be a convenience layer, a faster and more integrated way to access the AI assistant without living inside a browser tab. Instead, it’s become a case study in deprioritization dressed up as progress.
The Desktop App Was Actually Good — Here’s What Users Lost
Before OpenAI started retrofitting the ChatGPT desktop app to serve its enterprise ambitions, the Mac application had earned a genuine following among everyday users — not developers, not IT managers, just people who wanted a fast, reliable AI assistant sitting quietly in their dock.
The appeal was straightforward. The desktop app launched instantly, stayed out of the way, and delivered a tighter experience than opening yet another browser tab. For users who reached for ChatGPT dozens of times a day — to draft emails, summarize documents, think through decisions — the native app removed the small frictions that add up. No waiting for a browser window to load. No competing with 20 open tabs. Just the assistant, ready to go.
That use case is now broken. OpenAI’s push to roll out Codex for developers and ChatGPT Work for enterprise teams reshaped the desktop app around those audiences. Everyday features that casual users relied on were stripped out or buried in the process. The result is an application that feels half-finished for the people who originally made ChatGPT the fastest-growing consumer product in internet history.
Critically, this is a regression, not a tradeoff. Desktop apps exist for a specific reason: to do things a browser cannot. They run faster, integrate more deeply with the operating system, and offer a more focused environment. When a desktop app stops delivering on those advantages, users have no reason to install it. They go back to the browser — which is exactly what reviewers at ZDNET concluded after the changes took effect, noting that ChatGPT now works best in the browser for most users.
That sentence should sting for OpenAI’s product team. It means the company built, shipped, and then effectively abandoned a native application that millions of non-technical users had genuinely incorporated into their daily routines. The ChatGPT Mac app didn’t need to be all things to all people. It needed to stay good at the one thing it already did well.
Who Codex and ChatGPT Work Are Really For (Hint: Not You)
OpenAI built its reputation on the back of everyday users — students, writers, curious professionals, and people who just wanted a smarter search box. ChatGPT crossed 100 million users faster than any consumer app in history. Now the company is quietly pivoting away from those same people, and the product lineup makes that shift impossible to ignore.
Codex is OpenAI’s autonomous coding agent, designed to handle complex software engineering tasks inside a developer’s existing codebase. It operates independently, writes code, runs tests, and submits pull requests without constant human input. That is a powerful tool — for engineering teams at mid-size and large tech companies. For the teacher using ChatGPT to build lesson plans or the marketer drafting campaign copy, Codex is completely irrelevant.
ChatGPT Work follows the same logic. It targets business and enterprise customers, offering collaborative features built around organizational workflows rather than individual use. OpenAI is chasing the kind of high-margin, multi-seat contracts that companies like Salesforce and Microsoft have built entire revenue engines around. A business account paying per seat generates far more predictable, scalable revenue than millions of individual subscribers paying $20 a month.
The product roadmap now reflects that enterprise priority directly. OpenAI added agentic tools and enterprise-grade features to its core apps while stripping out capabilities that everyday ChatGPT users relied on. The desktop app, once a genuinely useful native experience for Mac users, degraded as the company reshuffled resources and attention toward Codex and Work. The browser version remains the most functional option for general consumers — which is a step backward, not forward, for a desktop app that once offered distinct advantages.
OpenAI is not the first AI company to discover that enterprise contracts pay better than consumer subscriptions. The strategic logic is sound from a business perspective. But the cost is real: the general-purpose AI assistant that millions of people adopted as a daily tool is now a secondary priority, quietly maintained while the company’s actual ambitions point somewhere else entirely.
The Missing Context: OpenAI Is Quietly Repositioning Itself
Most tech reporters covered the arrival of ChatGPT Codex and ChatGPT Work as straightforward product expansions — new tools added to a growing platform. Almost none asked what got quietly removed to make space for them. The answer is the desktop app experience that millions of everyday users had built their workflows around.
OpenAI is not simply adding enterprise features. The company is executing a structural pivot from consumer AI product to enterprise software platform. The degraded Mac desktop app is not a bug or an oversight. It is a symptom of a deliberate reallocation of engineering resources and product priorities toward developer tooling, agentic workflows, and corporate subscriptions — the segments that generate predictable, high-margin revenue.
This playbook is familiar. Google built its early dominance on free consumer products, then systematically wound down or neglected them once Google Cloud and Workspace became the real revenue engines. Microsoft followed the same arc, letting consumer-facing Windows features stagnate while pouring investment into Azure and Microsoft 365 Copilot for enterprise clients. OpenAI is now on the same trajectory, just moving faster.
The signals are clear. Codex targets software developers. ChatGPT Work targets corporate teams. OpenAI’s API business serves companies building products on top of GPT-4o and o3. Each of these audiences pays more per seat and signs longer contracts than individual ChatGPT Plus subscribers at $20 a month. When engineering bandwidth is finite, products serving those audiences win internal prioritization battles.
The consumer base that made ChatGPT the fastest-growing application in internet history — crossing 100 million users in two months — now occupies a lower position in the product hierarchy than it did two years ago. The desktop app degradation makes that hierarchy visible in a way that a press release never would.
What This Means for Everyday ChatGPT Users Right Now
If you use ChatGPT on a Mac or Windows desktop app and rely on it for daily tasks, your workflow is likely broken right now. OpenAI stripped everyday features from the desktop client to make room for Codex and ChatGPT Work — enterprise and developer tools that most casual users will never touch. The result is a ChatGPT desktop experience that is objectively worse than it was several months ago.
The browser version of ChatGPT remains the most reliable and feature-complete option for non-technical users. That is a regression. The desktop app existed precisely because it offered capabilities beyond what the browser provided — things like tighter system integration and faster access. OpenAI reversed that advantage and handed it back to the browser without a clear explanation to its general user base.
Anyone who built personal productivity workflows around the ChatGPT desktop app — scheduled tasks, document drafting routines, quick-capture habits — needs to audit those workflows immediately. Features that powered them may have been quietly removed or degraded. OpenAI did not send affected users a roadmap. It made the changes, and everyday ChatGPT users absorbed the disruption.
Anthropic is watching this closely — and moving fast. Claude, Anthropic’s AI assistant, recently launched Claude Cowork, a product explicitly designed for collaborative, everyday use cases. Where OpenAI is pulling resources toward enterprise clients and agentic developer tools, Anthropic is actively courting the general users that ChatGPT built its 100-million-plus user base on. That is a direct play for OpenAI’s most loyal and most neglected audience.
The practical advice for everyday ChatGPT users is blunt: switch to the browser, rebuild any broken workflows there, and start evaluating alternatives. Claude is the most direct competitor for users who want a polished, consumer-focused AI assistant experience. OpenAI’s recent decisions signal that the company no longer treats that user segment as a priority. Other AI platforms are ready to fill the gap.
Should OpenAI Have to Choose? The Case for a Separate Consumer Strategy
OpenAI does not face a binary choice here. Nothing in the technical architecture of ChatGPT forces the company to strip consumer desktop features every time it ships a new enterprise product. When OpenAI rolled out ChatGPT Work and Codex, the degradation of the Mac desktop app was a resourcing decision, not an engineering inevitability. The company chose where to point its product teams, and everyday users lost features they had come to rely on.
That choice carries a real cost that goes beyond inconvenience. The general-public users who downloaded the ChatGPT desktop app, recommended it to friends, and built daily habits around it were not beta testers — they were the foundation of ChatGPT’s cultural dominance. Pulling features from that experience without explanation, and without any public acknowledgment that trade-offs were being made, signals to those users that their needs are now an afterthought. Trust, once eroded at scale, does not recover quickly.
The fix is not complicated. OpenAI can run parallel product tracks — a dedicated consumer AI assistant experience optimized for everyday tasks like writing, research, and personal productivity, alongside a separate enterprise and developer platform built for agents, coding workflows, and business integrations. Other large software companies manage this split routinely. Microsoft maintains distinct consumer and commercial versions of core products. Google separates Workspace from personal Gmail. There is no reason OpenAI cannot apply the same logic to its AI chat software.
What the company cannot afford to do is continue treating the ChatGPT personal app as a secondary surface that absorbs collateral damage whenever a new business tool ships. The AI assistant market is no longer a one-player game. Claude, Gemini, and a growing field of specialized AI productivity tools are all competing for the same everyday users OpenAI is quietly deprioritizing. Losing that consumer base does not just hurt brand perception — it weakens the data flywheel, the word-of-mouth growth engine, and the long-term relevance of ChatGPT as a household name rather than just another enterprise software subscription.
