What's new in Tusk

Tusk now recalls granular insights from its memory of past closed and merged PRs as well as code reviews when completing new tasks. Previously, it only took inspiration from merged PRs, and it did so at a high level.

The more you use Tusk, the better it gets at generating code for your unique codebase.

Example of Memory in Action — Tusk using granular insights from memory to style a UI change appropriately

Other new releases

🛩️ Autopilot mode where Tusk auto-generates PRs for issues that it is confident in completing end to end (opt-in feature).

🧑‍🔬 Novel Mixture of LLMs approach that enhances Tusk's reasoning ability by combining the strengths of multiple SOTA LLMs.

🔔 Automatic reminders to review Tusk's PRs Tusk will auto-close its PRs if no action is taken after 10 working days.

⏪ Ability to view past attempts at PR generation and addressing code reviews for better visibility into Tusk's work.