- Edit the brief right where you read it. You can now click any section of the brief to change its wording in place — no separate form. Sections that shouldn't be touched are clearly locked, and removing a section asks you to confirm first, so nothing disappears by accident.
- MIRA speaks at your level. The assistant now reads how technical you are and matches it: name your stack and talk specs and she meets you there with the precise terms; keep it plain and she keeps it plain and explains anything specialized — never talking down to an expert or over a beginner's head.
- A smoother wait while talent is found. The search screen now flows as one continuous motion — the radar sweep settles straight into the row of talent instead of blinking over to it — and it holds its place properly. Refresh the page or come back later and the waiting screen rebuilds exactly where you were, on the right step.
- Picking a finalist actually reaches them. When you choose your finalist, that choice now travels all the way back to the system so STERLING can message the talent on your behalf — closing a gap where the pick could be made on screen but never acted on.
- The brief stops repeating itself. A timing issue where two quick back-to-back turns could write the same section twice — leaving a brief that said everything twice — is fixed: the assistant now builds the brief one turn at a time, so each part appears once.
- Quieter, sturdier matching. The assistant now reliably checks whether you need more than one person before it starts, the back-up search is seeded with the right keywords so it returns relevant talent, and the finalist reveal was trimmed so it always loads.
- A faster, safer engine room. The part of the system that needs a web browser was moved into its own sealed, isolated service instead of riding along in the shared one — cleaner and safer. Deploys were also made quicker and more reliable, so new work reaches the live site sooner and with fewer hiccups.
A polish-and-plumbing day: a brief you can edit in place, an assistant that matches how you talk, a calmer search that survives a refresh, and a quieter, faster engine room underneath.