A few recent changes make the content editor feel more predictable and under your control: you can edit HTML directly, stop a run that’s going nowhere, and see what the agent is doing while it works.

HTML Editor

Sometimes you want to tweak markup by hand — fix a class, adjust a tag, or paste a snippet. The new HTML editor lets you do that inside the chat interface.

You get syntax-aware editing and live preview, so you can switch between natural-language instructions and direct HTML without leaving the page. Useful for quick fixes and for learning how the AI structures the output.

Stop or Cancel Prompt Execution

Previously, once the agent started working, you waited until it finished. If the direction was wrong or you changed your mind, you had to sit through the rest of the run.

You can now stop or cancel a running prompt at any time. The run ends, and you can start a new instruction or refine your request. No more waiting out a long or off-track response.

Agent Progress Updates

When the agent is working, it now sends progress updates — short, chatty messages about what it’s doing step by step. You see “Reading the current file…”, “Updating the hero section…”, and so on, instead of a single “Working…” for minutes.

That makes it easier to tell whether the agent is on the right track and reduces the sense of a black box. Combined with the ability to stop or cancel, you have more control and visibility.

Agent Context Dump (For the Curious)

For debugging or documentation, you can trigger an agent context dump: the current system prompt, available tools, and conversation state. Helpful when you want to understand why the agent behaved a certain way or when you’re tuning project instructions.

In Short

These updates don’t change the core flow — describe what you want, get edits, preview, send to review — but they make the editor more flexible (HTML), more controllable (stop/cancel), and more transparent (progress updates and context dump). Small steps that add up to a better editing experience.