Rohit Malhotra
@rohit_malh5
Followers
92
Following
3K
Media
10
Statuses
142
Openhands Maintainer | Ex-CTO @sitewizai | NLP @ CMU | Primarily interested in Agents | Secondary interests in creative design
Joined July 2018
RT @rbren_dev: Nothing more frustrating than seeing "private scaffold" on public benchmark results. I love that model providers like Qwen a….
0
7
0
RT @Alibaba_Qwen: >>> Qwen3-Coder is here! ✅. We’re releasing Qwen3-Coder-480B-A35B-Instruct, our most powerful open agentic code model to….
0
1K
0
FWIW I think this applies to human developers too. I just see a lot of exceptions being made for agents under the guise of velocity.
0
0
1
I believe the scope of a task you delegate to an agent should match the scope you’re willing to review in a PR. IDE agents are tactical—you keep prompting and refine until the code looks good. But autonomous code agents like OpenHands are more hands-off. So how much work you.
1
0
2
OpenHands is so general-purpose that I now think of leveraging it with workflow-driven prompting. Also stating constraints works well for me. Examples:. • Examine the existing architecture, read docs for Y, plan how to implement X, then do it.→ Instead of: "Implement feature.
0
0
3
RT @MistralAI: Introducing Devstral Small and Medium 2507! This latest update offers improved performance and cost efficiency, perfectly su….
0
333
0
RT @gneubig: What will software development look like in 2026?. With coding agents rapidly improving, dev roles may look quite different. M….
0
16
0
PSA for engineering leadership exploring software agent solutions 🚨. This post nails the difference between agentic and agentless approaches — and why it actually matters for real software tasks, beyond SWE-Bench scores!.
Congratulations to Moonshot AI on their release of Kimi-Dev-72B, an open-weights model that achieves a great score of 60.4% on SWE-Bench Verified!. Our community tried it in OpenHands, but it didn't work well, only 17% accuracy. Is this surprising? Actually not really! 🧵
0
1
5
Some users click with code agents. Others struggle. Why? Agents are flexible and creative - just like their users! It's confusing!. Agents should understand, educate, and adapt to users. Even personalize. If the agent isn’t willing to grow, the user likely won’t either.
Agents like OpenHands are flexible and have common sense, adjusting to insensible or illogical user demands. Traditional software, however, follows rigid rules of behavior, which is then designed for imaginative users. "Good" design will change as software becomes more flexible.
0
0
1
We really do believe in open source, but it wouldn't be possible to sustain it unless someone took over these complexities. Shoutout @rbren_dev.
0
0
2
Open source is powerful. Building a company around it? Complicated. Strong insights on how to stay rooted in open source while juggling many stakeholder needs.
At @allhands_ai we are dedicated to:. 1. Making OpenHands the best agent for software development.2. Doing it in an open, community-driven way. To make this sustainable, we formed a company around the project. @rbren_dev explains our philosophy in our new blog.
1
0
3
RT @allhands_ai: What if we could have *trustworthy* agents that don't just write code, but also do research, understand multimodal content….
0
27
0
I believe “execution” and “evaluation” are two major challenges to adoption of code agents from a user perspective. Users must learn out how to leverage the agent effectively, and how to evaluate its work (asap). Could determine whether good agents also delivers great experiences.
0
0
2
Credits: This project was envisioned and led by.@poky_bear .
helenzhang.framer.website
Crafting interactive stories for physical and digital spaces to enrich embodied experiences.
0
0
2
Used OpenHands to build a speculative spatial tracking museum experience in Unity, despite no Unity experience. It was pure "vibe coding" with no tests or reviews. Challenges: adding features without breaking existing ones and managing a messy codebase. Review your code folks!
2
2
8
Agents like OpenHands are flexible and have common sense, adjusting to insensible or illogical user demands. Traditional software, however, follows rigid rules of behavior, which is then designed for imaginative users. "Good" design will change as software becomes more flexible.
0
1
2
When OpenHands gets confused about file locations or alters code unrelated to the (relatively simple) task, I now see it as a sign of excessive code complexity. I used to blame agent quality before, but more often than not, it's an issue of poor code quality.
0
0
2
Initial results look great 🚀.
We tested out Claude 4 in OpenHands, and yeah, it's good. 70.4% on SWE-Bench Verified out of the box, +10 points over 3.7 Sonnet. Our vibe checks also passed, so it's available in OpenHands and the OpenHands Cloud today.
0
0
3
RT @gneubig: Some people have said that OpenAI achieved state of the art results on the SWE-Bench Verified leaderboard with their codex mod….
0
28
0