More important and relevant in 2024

You can now apply SWE-agent to any local repository and use any text file as the input issue instead of having to use GitHub repos/issues. Lots of people were asking for this! More information in the latest release notes: github.com/princeton-nlp/…

Attention is all we need.
Memory is all we have.

It's been just 10 days since we launched SWE-agent but we already have 1.5k people in our Discord and lots of contributors on GitHub.

We've been making the agent easier to use and there are lots more exciting updates coming soon, including a web UI! Join us :)

🤔Can we assess agents across various apps & OS w.o. crafting new envs?

OSWorld🖥️: A unified, real computer env for multimodal agents to evaluate open-ended computer tasks with arbitrary apps and interfaces on Ubuntu, Windows, & macOS.

+ annotated 369 real-world computer tasks…

Guys, before achieving AGI, we need to solve zork-1?

Thanks Aran for sharing our work!

This is a survey paper I’ve been thinking about for a long time, as we have seen an increasing need for synthetic data. As we will probably run out of fresh tokens soon, the audience of this paper should be everyone who cares about AI progress.

Will visit AGI House for the first time this Saturday and talk about SWE-agent, Agent-Computer Interface (ACI), and answer questions😃

When I first saw Tree of Thoughts I also asked myself this😀 great exploration into if next-token prediction can simulate search, and if you're interested in this you probably also wanna check out arxiv.org/abs/2309.02427 last paragraph

You can now download & run SWE-agent (on any GitHub issue) in 1 line!

Check our repo for deets: github.com/princeton-nlp/…

Join our Discord to hear first about updates like this: discord.gg/AVEFbBn2rH

in some sense, math is the first programming language, and mathematician's mind (+scratchpad) is the first compiler

Purposeful pretraining and interface design may be the keys to unlocking reliable LM-based agents.

deliver like cole

People still surprised by such things across pairs among ReAct ToT Reflexion CoALA WebShop SWE-bench SWE-agent😂

Swe-agent achieves ~ Devin while only using < 2min and $4 per case

Fantastic work from our Princeton PLI team. Agent-based AI is clearly a big next step.

SWE-agent is finally out. A few highlights:
1. Agent-Computer Interface (ACI) design will be critical for the success of AI agents, much like HCI is critical for how effective humans are with computers.
2. You can use SWE-agent out of the box on any github issue.
(1/2)

What we call reasoning in AI is algorithm in CS

A neat idea and setup

People are asking us how Claude 3 does with SWE-agent- not well. On SWE-bench Lite (a 10% subset of the test set) it gets almost 6% less (absolute) than GPT-4.

It's also much slower.

We'll have all the data in the preprint next week.