Soft ComputersWhat OpenAI Actually Did
OpenAI recently pulled several features from its ChatGPT desktop app. According to ZDNet, the company removed capabilities from the desktop experience specifically to clear the way for two new products: Codex, an AI coding agent, and a separate tool called Work. The headline from a ZDNet writer who had been using the app daily said it plainly: they loved the desktop app until OpenAI gutted it.
This is not a minor update. Features that people had built workflows around were simply gone. And the reason was not a bug. It was a product decision to shift focus toward developer-facing AI tools, with Codex at the centre.
What Is Codex, and Why Should You Care
Codex is OpenAI's AI agent designed specifically for writing, reviewing, and running code. It is not just an autocomplete tool. It can take a description of a task and work through multiple steps to produce functioning software output. OpenAI is positioning it as something closer to an autonomous junior developer than a chat assistant.
For Canadian SMBs that are building or maintaining software, websites, or internal apps, this is worth paying attention to. If your team has been using ChatGPT casually to get coding help, the product you were relying on has changed. And if you have not been using AI for development work at all, Codex represents a meaningful shift in what is actually possible for smaller teams without large development budgets.
The Real Problem: Tools That Change Without Warning
Here is the part that catches businesses off guard. A team adopts a tool, builds a routine around it, and then the vendor changes the product to serve a different market. That is exactly what happened here. OpenAI decided Codex and Work were higher priorities, and the desktop app users absorbed the cost of that decision.
We see this pattern regularly. A business starts relying on a SaaS tool or an AI assistant for real work. No formal evaluation, no documentation of what the tool is doing, no plan B. Then the vendor pivots, raises prices, or removes features, and the disruption hits a team that was not ready for it.
This is not an argument against using AI tools. It is an argument for using them deliberately.
What We Recommend for SMBs Using AI in Development
If your business is using AI tools like ChatGPT, GitHub Copilot, or similar products to support software or web development work, a few things are worth doing now.
- Audit what your team actually depends on. If someone on your team is using a specific AI feature to do real work, that dependency should be visible. It should not live only in one person's browser session.
- Know which tier you are on. OpenAI's Codex access is tied to specific ChatGPT plans, and pricing and availability have been shifting. Check whether your current subscription includes access to the tools you are counting on. As of mid-2026, Codex is being rolled out through ChatGPT Pro and Team plans, but availability and limits are subject to change.
- Do not treat AI as a replacement for a development process. Codex and similar agents can accelerate real work, but they need oversight. Code produced by an AI agent still needs review before it touches your website, your customer data, or any system your business depends on. We have seen AI-generated code pass a quick look and still introduce security gaps.
- Evaluate GitHub Copilot as a parallel option. For teams doing regular software or web development work, GitHub Copilot (now at version Copilot Enterprise for larger teams, or Copilot Individual at $10 USD per month as of this writing) integrates directly into code editors like VS Code. It has not been subject to the same kind of feature removal that just hit the ChatGPT desktop app, and its roadmap is more explicitly developer-focused.
If You Are Thinking About a Custom Software or Web Project
The emergence of tools like Codex does change the economics of custom development somewhat. AI-assisted development is real, and when used properly it can reduce the time and cost of building something from scratch. But the tooling is still changing fast, and the gap between a working prototype and a production-ready application is still significant.
Our development team uses AI coding tools as part of the build process, not as a substitute for architecture decisions, security reviews, or quality testing. If you are evaluating whether to build a custom app, a web platform, or an internal tool for your business, we are happy to walk through what a realistic scope and timeline looks like given how the tooling has evolved.
The short version: AI is making some things faster and cheaper. It is not making the judgment calls easier.
Bottom Line
OpenAI's decision to strip down the ChatGPT desktop app to prioritize Codex is a reminder that AI products are still moving fast and vendor decisions do not always align with what your business has built around them. Stay deliberate about which tools you depend on, know what plan you are on, and make sure any AI-assisted code gets a proper review before it goes anywhere near a live system.
If you have questions about how to use AI tools responsibly in a software or web project, reach out to our team. We are based in Toronto and work with Canadian SMBs across a range of industries.