Juan Michelini retweetledi
日経ソフトウエア 2026年3月号
雑誌 – 2026/1/23
amzn.to/3Z1Oxnv amazon #ad
AI エディタ入門
Visual Studio Code
GitHub Copilot
プログラミング向け AIエージェント
OpenHands 自然言語でアプリを作る
生成AIコーディング道場
日本語
Juan Michelini
13 posts
Juan Michelini
@juanontech
I love when numbers go up 🌊 AI Agent Researcher @openhandsdev 🤖
Katılım Mart 2025
34 Takip Edilen35 Takipçiler
my IDE? github issues #issuecomment-3498861079" target="_blank" rel="nofollow noopener">github.com/OpenHands/benc…
English
Juan Michelini retweetledi
Juan Michelini retweetledi
3/ This week we are proud to announce our latest eval with the latest version of @allhands_ai's OpenHands and GPT-5.
Remarkably, GPT-5 increased the SOTA on SWT-Verified by 15%, taking 1st, and GPT-5-mini took a well deserved 3rd spot. 🏅
English
Juan Michelini retweetledi
A lot of agents out there are over-optimizing for SWE-bench. We've been very careful to ensure OpenHands generalizes to a wide variety of eng-related tasks
Great to see that work pay off as we hit #1 on SWT-Bench!🏆
OpenHands@OpenHandsDev
Having appropriate tests makes a world of difference for agent-driven development. If your agent can write a test to localize a bug or exercise a new feature, the following implementation is much more solid. OpenHands+GPT-5 is now 🥇 on the SWT-Bench testing leaderboard!
English
OpenHands is 🥇 place on the SWTBench the unit test benchmark. Great team, great product, and the benchmarks prove it!
OpenHands@OpenHandsDev
Having appropriate tests makes a world of difference for agent-driven development. If your agent can write a test to localize a bug or exercise a new feature, the following implementation is much more solid. OpenHands+GPT-5 is now 🥇 on the SWT-Bench testing leaderboard!
English
Juan Michelini retweetledi
Having appropriate tests makes a world of difference for agent-driven development.
If your agent can write a test to localize a bug or exercise a new feature, the following implementation is much more solid.
OpenHands+GPT-5 is now 🥇 on the SWT-Bench testing leaderboard!
English
Juan Michelini retweetledi
We evaluated GPT-5 in OpenHands and it's the new number one coding agent model for us!
Using exactly the same tools and harness it's 1.4 points better than Claude Sonnet 4 at 60% of the price.
Full results here: #gid=0" target="_blank" rel="nofollow noopener">docs.google.com/spreadsheets/d…
English
Juan Michelini retweetledi
@MistralAI Try it out in OpenHands following the directions here! docs.all-hands.dev/usage/llms/loc…
English
Juan Michelini retweetledi
Introducing the OpenHands CLI, a new coding CLI that:
- Has top accuracy (similar to Claude Code)
- Is completely open source, MIT licensed
- Is model agnostic, use an API or bring your own
- Is simple to install and run `pip install openhands-ai` and `openhands` (no Docker!)
English
@engelnyst Yes! Software Engineering is the first thing we automate because it's the one we know the best. Also, that's where all the exponential growth comes from.
English
We've always tried to automate ourselves out of a job!
When I was building Eclipse plugins and RCPs, I thought, must-must automate writing code
When building the conversion of UML diagrams into code, of course it was going to finally make it
It's that time again 😂
Klaas@forgebitz
how tf did we automate software engineers before real estate agents this is such a big fumble on our end
English
English
Huge shout-out to @Aditya_Soni_8 at CMU, who's amazing work on his paper laid the foundation for accuracy improvements on many tasks: x.com/Aditya_Soni_8/…
And Juan at @allhands_ai, who set up VersaBench to do such a diverse variety of benchmarking.
Aditya Soni@Aditya_Soni_8
Can we design AI Agents that achieve generalizability across diverse task domains? Our new paper introduces OpenHands-Versa, a generalist agent with strong performance on three challenging agent benchmarks, ranking #1 on SWE-Bench Multimodal and The Agent Company leaderboards 🚀
English
Juan Michelini retweetledi
I am very excited about both the accuracy results and just my feeling playing around with this new agent.
The ability to research things online and comprehend a wide variety of multimodal content has removed many of my blockers in using agents in my everyday work.
OpenHands@OpenHandsDev
What if we could have *trustworthy* agents that don't just write code, but also do research, understand multimodal content, and perform many practically useful tasks? Today at OpenHands, we released a new agent that gets SOTA or competitive performance on 8 diverse tasks.
English