Prompt Analyzer: Iterate, Improve & Save Versions

Version: 1.28.0, Date: June 4, 2026

Prompt analyzer iteration is the theme of this release. The Prompt Analyzer is no longer just a place to grade a prompt and read a static report: you can now analyze, refine with one-click AI actions, watch each pass produce a full comparable result, and save the whole lineage back to your Library when you land on a version you like.

Highlights

• Iterative improvement loop: analyze, refine, and re-analyze in tracked rounds without leaving the page.
• One-click improvement actions: Improve, Strengthen, Transform, and Adapt open into roughly two dozen targeted sub-actions.
• Save as versioned lineage: keep the original prompt and every improved pass together in your Library.
• Plain-language Insights: every round explains what changed, whether scores moved, and what to try next.

Analyzers

• New: Iterative improvement loop. Analyze a prompt, then keep refining it. Each improvement runs a fresh analysis automatically and creates a new round you can flip back to, so the full history of how a prompt evolved stays in front of you.
• New: Five views per round. Every round exposes Analysis, Prompt, Insights, Details, and Ratings as separate tabs. When you switch between rounds your active tab stays put, making it easy to compare the same view across versions.
• New: Two ways to load a prompt. Start from a prompt you paste in, or pull one directly from your Prompt Library to analyze and improve.
• New: One-click improvement actions. Improve, Strengthen, Transform, and Adapt open into roughly two dozen targeted sub-actions, alongside a Reanalyze option and an “Apply insight recommendations” action. Each one rewrites your prompt toward a specific goal and immediately re-analyzes the result, naming the new round after the action you ran.
• New: Plain-language Insights. Beyond scores, the analyzer now coaches you. The first round gives a verdict with strengths and weaknesses; every improvement round explains what changed, whether your scores moved, and what to try next.
• New: Comparable Ratings across rounds. Ratings show current-versus-recommended bars for each dimension, and the dimension names stay consistent across an improvement chain so you can genuinely compare scores from one round to the next.
• New: Save versions to your Library. When you save, the analyzer keeps your work as a versioned lineage: your starting prompt becomes the original, and each improved pass is saved as a version beneath it. If you started from a library prompt, the improvements are added as new versions of that source, so the refinement history lives with the prompt itself.
• New: Shared content toolbar. The Analysis and Prompt views now share a consistent toolbar: expand or collapse, edit inline, toggle between formatted and raw text, copy, and export to HTML, Markdown, or plain text.
• New: Details view with consumption visibility. Each round includes a Details card with run info, the prompt’s source, and your AI consumption for that round. Consumption is shown as tokens if you are using your own key or credits if you are on platform billing, totaled across every call in the round.
• New: Stop control. If a request hangs, you can now cancel it on the spot instead of waiting it out.

• Improved: More reliable Ratings and Insights. Ratings and insights are now generated through a dedicated step, so they no longer get cut off or skipped when an analysis runs long. If something does fail, a Retry control is right there on the tab.
• Improved: Accurate consumption totals. The completion notification now appears once the entire process finishes and shows the full combined token or credit total for the round, matching exactly what you see in the Details view.
• Improved: Refreshed Analyzer help documentation that walks through the new iterate-and-save experience end to end.

• Fixed: Out-of-credits is no longer silent. If you run out of credits while insights are generating in the background, you will now see a clear notice with a retry option and your quota details, instead of the step quietly failing.
• Fixed: No more duplicate rounds. Rapid double-clicks can no longer kick off duplicate analysis rounds.