Claude Opus 4.7 launched. I'm fixing the AI workspace so it doesn't lose your place.

Anthropic launched Claude Opus 4.7 in the last 24 hours. The model layer keeps moving. Inside the product, the pressure is simpler. 186 merged events across Claude Code, Codex, and Letta. A lot of motion. Same underlying problem. The workspace is still losing your place. I went to sleep and came back to dropped window sessions. One survived. Two did not. Terminal tab names still need to persist. The command palette still activates every window when it should stay inside the current pane. Dragging a file into the terminal still opens the file instead of pasting the path. The topology view can draw a graph, but the system still needs to explain itself cleanly. The messaging area is trying to become an artifact surface, not just a text log. Those were not separate bugs. They were different ways of finding the same gap. The room still forgets where you were, what pane you were in, and what should open where. Opus 4.7 is the outside signal. Losing your place is the inside one. If the workspace cannot hold state, survive sleep, and open the right thing in the right place, the whole agent story falls back into demo mode. So this stream is the keep-your-place pass. Pane behavior. Session resilience. Artifact previews. Topology clarity. The parts that decide whether this is a place you visit or a place you can actually work from. If you code, you probably know this kind of session. The model is not the blocker. The operating surface is. AI radio on. Quiet co-working. 8 to 10 hours of making the workspace stop losing your place. Anthropic: https://www.anthropic.com/news/claude-opus-4-7