The long-awaited OpenAI "potato" dropped today. Turns out it's not GPT-6, but GPT-5.5. I tried it out right after it came out and found that it fixed the thing that annoyed me most about 5.4: it finally started talking like a human . Back in the 5.4 days, reading through Codex's output process was torture. Words piled on top of each other with no logic, status reports and execution notes jumping back and forth—you mostly couldn't tell what it was actually doing. 5.5 fixed this, and the readability of the entire workflow shot up immediately. And then it gets funny. Just a while ago I was telling my team: GPT 5.4 is stronger than Claude on certain complex tasks, but it doesn't talk like a human, so it's exhausting to use; Claude speaks human, it can guide you through, so we should still push Claude Code . Some colleagues went and canceled Codex and switched to Claude after hearing that. Today I have to send another message in the group chat: maybe just get both accounts and use them together.…