Tag Selection Simulation, Highlights Are Now Facts, Not Insights + 5 more
Tag Selection Simulation
Another week, another set of meaningful improvements. Here is a quick, high-signal roundup of what is new, focused on clarity, speed, and trust in every step of the workflow.
We added Tag Simulation Mode: a safe sandbox where you can test tag descriptions, prompts, and context without touching production data. Experiment freely, see how the system would behave, and only apply changes once you are confident. Fewer surprises, better decisions.

Highlights Are Now Facts, Not Insights
Highlights have been reframed from interpretations to literal facts. This removes assumptions and keeps outputs strictly grounded in the source material.
- More neutral and consistent highlights
- Cleaner inputs for clustering, tagging, and UX analysis

Faster and Smarter Clustering
Our clustering algorithm received a substantial upgrade:
- 20–50% faster runtimes
- Higher match quality with fewer false positives
Clearer Thinking Feedback in Chat
We replaced the old pulsating dot with a subtle animation on the status text, such as “Generating clusters…”. As long as it is animating, you know the agent is actively working, making long-running steps feel transparent instead of stuck.
Much Faster Search Suggestions
Search suggestions are now up to 10× faster, especially when refining queries. Less waiting, more flow.
Email Alerts When Imports Stop
If an integration subscription becomes disabled, for example due to expired or revoked access, we now send an email immediately explaining what happened and what to do next. There are no more silent failures or detective work weeks later—just immediate visibility and a direct link back to settings.

Goodbye Prompt-Generated Dashboards
We removed the option to create a dashboard from a prompt. These dashboards often came with confusing, opaque filters. Every new dashboard now starts as a clean slate, with no hidden logic or mystery setup, keeping you fully in control.