Previously, your credit balance would only reflect after a page refresh — meaning after running an extraction or prefill you’d see the old number until you reloaded. The balance now updates immediately after each AI action, so you always have an accurate view of what you’ve used.
Reduced the average end-to-end response time for both outline extraction and submission prefill. The improvements come from better request batching on the backend — no change to the quality or depth of output, just less time waiting.
Plan upgrades and credit top-ups were sometimes taking several seconds to reflect, with the UI left in an ambiguous state. Billing changes now apply synchronously with clear confirmation feedback — when the confirmation appears, the credits or plan change is already active.
Workspaces with a high number of projects were experiencing noticeable slowdowns when navigating between views. Resolved the underlying query bottleneck — large workspaces now load and navigate at the same speed as smaller ones.