Sabitlenmiş Tweet
Cua
662 posts
Cua
@trycua
Open-source infrastructure for Computer-Use Agents // YC X25
San Francisco, CA Katılım Ocak 2025
1.8K Takip Edilen4.9K Takipçiler
Cua retweetledi
English
Cua retweetledi
Cua retweetledi
Excited to see our research partners at Snorkel AI launch Open Benchmarks Grants - $3M to close the evaluation gap in agentic AI. This is one of the biggest bottlenecks in the space and it's finally getting funded properly. Apply → benchmarks.snorkel.ai/apply
vincent sunn chen@vincentsunnchen
Our ability to measure AI has been outpaced by our ability to develop it, and this evaluation gap is one of the most important problems in AI. Today we're launching Open Benchmarks Grants — a $3M commitment to fund open benchmarks for frontier AI and close the evaluation gap. Grateful to be partnering with @HuggingFace, @togethercompute, @PrimeIntellect, Factory HQ, @harborframework, and @PyTorch to back the teams building these benchmarks! 🚀
English
Cua retweetledi
Yesterday we demoed the very first multi-player computer-use agent at @openclaw ClawCon, in front of @steipete @aplusk @RayFernando1337 @theo @swyx and a 700+ crowd with 20k watching online.
Read more about how and why we built it.
Cua@trycua
English
Cua retweetledi
Cua retweetledi
Make your agents better at computers.
Cua-Bench evaluates computer-use agents on real tasks across desktop and mobile platforms.
English
Cua retweetledi
I keep saying this...
2026 is going to be very interesting for computer-use agents.
a lot of folks are building with these frameworks right now.
that means we're going to see the products they're building launch in the coming months... 👀
Francesco@francedot
Trending on GitHub again in Python, right next to our friends at @Prince_Canuma - thank you for 12k ⭐ - your support means everything. Chef @trycua cuala will keep cooking releases this week👨🍳🐨
English
Cua retweetledi
Love to see the ecosystem evolving around clawdbot. Very excited by what CUA is doing here, excited to try this out later today.
Francesco@francedot
English
Cua retweetledi
my calendar is now open for a @trycua / lume collison installation / if you're in SF i'll personally come to your office: cal.com/cua/cua-collis…
Ivan Fioravanti ᯅ@ivanfioravanti
Solution to run multiple clawdbots on my Mac Studio found: lume! 🔥 Gonna try later today!
English
Cua retweetledi
Cua retweetledi
This is why I encourage anyone to try a startup, only @francedot could push me to try something trending on X/Twt after he saw it 2 h ago and make it production ready + actually ship it!
Result? We did 130k+ impressions on our blog post and video and we open sourced it and we continue to work on computer use :)
@trycua gratitude!!
Francesco@francedot
A bunch of you asked about our Remotion setup after the article. It's now open-source: github.com/trycua/launchp… • Video templates for product launches • Shared animation components • Works with Claude Code + Remotion skills • How we made the Cua-Bench video in 2 hours
English
Cua retweetledi
"recording a human demonstration via noVNC, turning it into a Claude skill, then a CUA agent replaying it on the same spreadsheets task from the ShowUI-Aloha paper"
Cua@trycua
English
Cua retweetledi
Computer use agents that can learn, remember, and reapply your workflows on the fly! Cua has been shipping awesome stuff this entire week, check it out!
Cua@trycua
English
Cua retweetledi
Cua retweetledi
Congrats to @trycua on open-sourcing cua-bench!
We're collaborating on task design & data curation for GUI workflows - bringing the same systematic evaluation approach from Terminal-Bench to computer-use agents.
15 native tasks / 40 variations + OSWorld & Windows Agent Arena.
Early cua-bench-basic results:
-> Opus 4.5: 83.8% (57/68)
-> Sonnet 4.5: 79.4% (54/68)
-> Haiku 4.5: 79.4% (54/68)
These are basic tasks — stay tuned for much more complex, realistic workflows 👀
Cross-platform. MIT licensed: cua.ai/docs/cuabench
Cua@trycua
We've been using Cua-Bench internally—and with customers—for the last few months to evaluate every computer-use agent we deploy. Today it's open-source. 15 public tasks, 40 variations, adapters for OSWorld and Windows Agent Arena. One CLI, self-hostable.
English
Cua retweetledi
One-stop cua evaluation, seems amazing!
Cua@trycua
We've been using Cua-Bench internally—and with customers—for the last few months to evaluate every computer-use agent we deploy. Today it's open-source. 15 public tasks, 40 variations, adapters for OSWorld and Windows Agent Arena. One CLI, self-hostable.
English