06/27/2024

Thread

Tweet 1

For the last few months I’ve been working at Abante on a github bot that solves issues and reviews PRs. Today it’s generally available and we’re announcing we benchmarked 38.3% on SWE-bench lite

---

Tweet 2

I made this tweet the day we got SWE-bench setup and actually solved the first issue on the first try and then 3 out of the first 5. That turned out not to be representative. But we still got good results.

---

Tweet 3

You can read more about our approach and results here

---

Tweet 4

And you can try it yourself on your repos today here

---