Tusk

New Tusk landing page just dropped. Our goal has always been to save developers time. One thing that has repeatedly come up is the pain of writing high-impact tests. Which is why we're going all in on helping developers generate unit and integration tests for PRs. Tusk uses codebase context along with the optimal mix of LLM-based and programmatic approaches to catch edge cases your engineering team hasn't covered. This works for frontend and backend tests. If you're curious, reach out to Sohil, Jun Yu, Sohan or me to learn more.

This spooky season, we've committed as always to moving scary fast. Other elephants plod, this one sprints. ????????'?? ?????? ???? ???????? ? In October, we've introduced the ability for Tusk to sample from multiple parallel runs for the ideal path, better match Figma components to your repo's components, released Automated Browser Testing to the wider public, and made optimizations to both code exploration and generation. ?????? ???????????????? ?? - Ability to determine best path from multiple parallel agent runs - Figma integration revamp to better find matching components - Automated Browser Testing optimizations and self-healing - LLM pre-run warnings for complex issues - Support for new Claude 3.5 Sonnet - Enhancements to codebase exploration and recursive agent - Slack app to notify teams of merged Tusk PRs - Improvements to code generation for large files using AST - Ability to replicate any run’s output to a new PR - UI/UX improvements to Dashboard and Activity Log to better manage issues - Updated marketing website

New customer story out! Here's how Genie ?? completed 1 in 4 feature tickets using Tusk ??, our AI coding agent. Genie faced the classic challenge of having to build a compound product from the outset being in the e-commerce space. Their product team would receive a constant stream of customer feedback, which over time created a large backlog of small to mid-sized feature requests. Sebastiaan Debrouwere, Founder & CEO, needed a way to mow down this backlog quickly to make customers happy. However, Genie had?limited engineering bandwidth?for completing these customer-requested features. He turned to Tusk to help scope and complete these features. Our AI agent allowed him and the engineering team to go from ticket to merged PR in hours instead of days or weeks. Learn more in the comments about how they saw 4x ROI using Tusk. ?? Special thanks to Sebastiaan Debrouwere and Jimin Park for the continued collaboration and energising brainstorming sessions. ?? Sohil Kshirsagar Jun Yu Tan Sohan Kshirsagar

Shh, just pretend it's still last month. Anyway, here's our Sep 2024 changelog. ????????'?? ?????? ???? ???????? ? In September, we launched the beta of?Automated Browser Testing, added support for OpenAI's newly-released?o1 models, and shipped features to improve Tusk's code generation quality and overall user experience. ?????? ???????????????? ?? - Automated Browser Testing beta - OpenAI o1 support - Better guardrails for different task types - Improved UX for comments in ticketing apps - Introduced rich-text editor to description fields - UI/UX improvements to Dashboard, Suggested Issues, and Settings pages

It's being a game changer. To use Tusk.

Excited to announce that Tusk is partnering with Momentic to bring AI testing superpowers to our AI coding agent. Today we’re releasing Automated Browser Testing. Tusk is now the first AI coding agent capable of autonomously testing its frontend changes and iterating on them like a human engineer would. Automated Browser Testing builds on top of Momentic’s SOTA web agent. We generate testing steps in natural language and replicate user flows live in a browser on-demand. We’re excited to partner with the Momentic team for this key milestone. Get the power of Tusk’s code generation with the peace of mind that every ticket is validated in a real, live browser. Read more: https://lnkd.in/gdqPFXnd #AI #LLM #Partnership Sohil Kshirsagar Jun Yu Tan Wei-Wei W. Jeff An Henry Haefliger

New Tusk blog is up! ?????? Stay in the loop with our latest feature releases, technical deep dives, and other news. Our team will also regularly write about engineering-first stories and learnings that we personally find interesting in the LLM, autonomous agents, and software development space. Subscribe to the newsletter. Link to the blog in the comments ??

We're launching something big for Tusk (YC W24) soon. Big enough to give us a step change improvement in agent reliability. And we're looking for beta users. Without spoiling the surprise, we've created a way for our coding agent to emulate how an engineer would test their frontend changes. If you're an engineer or PM on a team that deploys preview environments on every pull request (e.g., Vercel preview links), reach out to me in the comments or over DM and I can provide beta access if it's a fit. Sohil Kshirsagar Jun Yu Tan

Tusk was #1 SaaS Product of the Week on Product Hunt the week of Aug 12. Thanks all for your support! ???? Improving agent reliability, steerability, and onboarding has been on our minds. Changelog for the past month: - Recursive Agent to improve reasoning - Abstract syntax tree to improve code generation - Custom repo instructions exposed in the UI - Suggested Issues from project management app - Improvements to addressing automated checks - Ability to retry code generation with feedback - Cleaner onboarding flow - Improved issues table UX

We're live and on the front page of Product Hunt! ???? Go to the Product Hunt home page and support our launch. We've got a good community of folks asking thoughtful questions and providing interesting feedback in the launch post. Check out the Product Hunt link in the comments ?