A ChatGPT user switched to Claude for a week.
He was impressed by the writing quality the responses read like a thoughtful colleague instead of a machine. The code was cleaner on the first draft. The analysis had nuance he wasn't used to. But he couldn't find the features he relied on. "Where's computer use? Where's Sites? Where's the browsing? Where's the live website builder?" He concluded Claude is "a great writer but limited" and went back to ChatGPT.
A Claude user switched to ChatGPT for a week.
She was impressed by the computer use the AI opened browsers, navigated websites, and completed multi-step tasks while she watched. Sites built a live landing page from one paragraph. The research was autonomous instead of collaborative. But the responses felt different. "The answers are faster but shallower. The code works but needs more follow-up fixes. The writing is competent but sounds like AI." She concluded ChatGPT is "a capable assistant but less thoughtful" and went back to Claude.
Both conclusions are wrong.
Both users tested the surface of the unfamiliar platform for 7 days and judged the entire product on the 20% they discovered without finding the hidden capabilities that make each platform extraordinary at tasks the other platform's users assume it can't handle.
The ChatGPT user who called Claude "limited" never found Claude Code, extended thinking, the 1-million-token context window, or the effort adjustment that turns a fast answer into extraordinary analysis. The Claude user who called ChatGPT "shallower" never found recurrent depth, reasoning effort adjustment, Codex persistent notes, or the prompting techniques that extract the deep thinking Astra is capable of but doesn't apply by default.
Here's every hidden capability on both platforms that the other platform's users never discover 🧵