OpenAI’s Mac app reshuffle has turned into a product-design argument about what ChatGPT is supposed to be: a straightforward chat interface or a more complicated workbench descended from Codex.
MG Siegler, writing at Spyglass, said OpenAI appears to be moving quickly to repair the rollout after a stronger-than-expected user reaction. His read was that OpenAI misjudged how people use ChatGPT, even if the company may have expected some complaints.
Siegler said the company has a harder problem than most rivals would complain about: a very large ChatGPT user base. He described that as an asset and a product constraint. If OpenAI wants to change the app’s identity, it has to do that without breaking the main thing users came for.
His narrower complaint was simpler. He wanted the chat interface restored to the front of the ChatGPT app. Siegler wrote that OpenAI has now done that, though he still called the app bloated and imperfect. In his view, the updated app is at least usable again rather than a jumble of competing ideas.
Daring Fireball was less forgiving. The site argued that OpenAI’s original split made sense: ChatGPT was the simple app, while Codex was the more complex tool. Adding a ChatGPT tab to Codex would have been fine, it said. Renaming Codex as ChatGPT was the bad decision.
The complaint is partly about naming and partly about workflow. Daring Fireball said the new Mac app now called ChatGPT lacks features that made the older ChatGPT app better for ordinary chat use. One cited example: deleting a chat reportedly requires archiving it first, then finding it in the archive and deleting it there.
The naming is also messy across platforms, according to the post. On iOS and Android, the ChatGPT app remains the older, simpler ChatGPT. On the Mac, that older app is now called “ChatGPT Classic,” while a more complicated app carries the main ChatGPT name.
Daring Fireball’s proposed fix is blunt: restore the old names. Call ChatGPT “ChatGPT” and Codex “Codex,” or, if OpenAI wants to connect the brands, rename Codex to “ChatGPT Codex” rather than demoting the existing app to “Classic.”
The larger point is that product complexity has consequences. Daring Fireball cited OpenAI’s own accounting that ChatGPT has about a billion users while Codex has only a few million. Its argument is that ChatGPT’s reach comes from being focused on one coherent idea, chat, while Codex is aimed at a narrower and more technical audience.
The criticism lands on OpenAI’s product judgment, not the underlying capability of Codex. Daring Fireball said Codex can do impressive work. The objection is that OpenAI tried to solve adoption by moving the ChatGPT label onto a different kind of app, then partially backfilled chat after users objected.
This story draws on original reporting from Daring Fireball.
