ENGINEERING NOTES · 9 NOVEMBER 2026
Less noise.
More evidence.
We want Yas apps to feel light and responsive. Here is what we can currently say, and how we plan to measure progress.
What is implemented
| Fact | What it means |
|---|---|
| Rust desktop core in both apps | Yas Mail uses Tauri 2. Yas Browser uses Tao and Wry. Their UI and web content also use web technologies. |
| Windows WebView2 rendering | The apps use Microsoft’s runtime. Its processes still use memory and CPU and must be included in measurements. |
| Yas Browser lazy session restore | Saved tabs can remain as lightweight entries until opened. The browser does not need to load every restored website immediately. |
| Yas Browser sleeping tabs | Eligible background website views unload after two minutes. Active or paused downloads can keep a source view awake. Reloading a sleeping tab can lose unsaved page state. |
| 20-tab local resource check | The v0.4.5 report checks retained tab URLs and website-view residency using synthetic local pages. It is not a Chrome/Edge comparison. |
What we are not claiming yet
We have not established a general percentage reduction in RAM, CPU, battery use, or overall device load against Chrome, Edge, Outlook, Mailbird, or other clients. In particular, a “70% less” claim would be premature. Rust is an implementation choice, not evidence of a specific performance advantage.
How a fair comparison should work
- Use the same device, operating system, power settings, and app versions.
- Include the app and all renderer/runtime child processes.
- Use the same websites or mail workload, account data, tab counts, and activity sequence.
- Measure startup, active work, and idle periods separately; report CPU over a stated interval.
- Repeat runs and publish the individual results, median, memory metric, and test conditions.
- Explain feature differences—especially sleeping tabs, unloaded pages, and background sync.
A smaller installer does not equal lower RAM use. An unloaded tab does not do the same work as a live page. If a comparison shows a real reduction, we will label the exact metric and workload rather than turn one result into a universal promise.
Try the Windows previews ↓