San Francisco Bay Area
Joined February 2022
- Tweets937
- Following6
- Followers9.7K
- Likes913
Simular retweeted
🪄 If your computer cursor is moving on its own, make sure it’s actually getting work done.
Get your halloween Saibot: sai.simular.ai
Simular retweeted
It’s not revealed in the paper but this is fundamentally learning a world model of the digital world. This paper marks a new era of computer use agents.
Our agent scientist, Agent S4, just got an open 27B model to beat GPT-5.5 and Claude Opus 4.8 on OSWorld V2, with the model's weights left exactly as they were.
With the interfaces S4 evolved, Qwen3.8-27B completes 21.3% of the 108 tasks, the highest binary success of any open-source computer-use agent and 3.8x the same model with fixed interfaces.
It costs $1.10 per task, against $25.50 for GPT-5.5 and $72.40 for Opus 4.8 💸
Full write-up: simular.ai/articles/agent-s4…
Simular retweeted
CUA should back. I once tried Astra's CUA to get me a car insurance quote 🤣. It hit the usage limit before finishing, and burned money at lightning speed. And our Opus 5.5 doesn't seem great at CUA either. There's definitely room here, and a research direction: a model built for long, unstructured prefill input + fast, cheap decode output.
👀 Agent S4 make CUA not really like CUA, it's more like a combined one. Agent will have a low level action (click scroll, and type) but could also come up with better interface to packing some low level action together (and combine with CLI). I think it makes sense and also could potentially benefit all the vertical agent, it could just make the thing works better and cheaper.
(I try CUA in Lightroom with my photo, save my time :)
Our agent scientist, Agent S4, just got an open 27B model to beat GPT-5.5 and Claude Opus 4.8 on OSWorld V2, with the model's weights left exactly as they were.
With the interfaces S4 evolved, Qwen3.8-27B completes 21.3% of the 108 tasks, the highest binary success of any open-source computer-use agent and 3.8x the same model with fixed interfaces.
It costs $1.10 per task, against $25.50 for GPT-5.5 and $72.40 for Opus 4.8 💸
Full write-up: simular.ai/articles/agent-s4…
Simular retweeted
@Richard_simular presenting neuro-symbolic computer use agents at @BayLearnOrg today!
Simular retweeted
Build your own personal agent within 10 minutes, using the Autonomous Computer API by @SimularAI
It also self optimizes token usage for repetitive tasks btw
youtu.be/doddTcFZMw8?si=YQWP…
Simular retweeted
Writing questions for our guests tonight 🪭 we have a full house already in the Mission. Let me know if you still want in! #SFTechWeek
featuring @reiinakano @LytvynovaYuliia @AJMadani and Gage from @frontiertower
partiful.com/e/YUKBbGuv0GuGM…
Simular retweeted
Sai Enterprise integrates @SimularAI’s state-of-the-art computer use with @Microsoft MXC, bringing guardrails to shell commands on local and cloud computers.
Watch Sai in action!
Excited to be a partner for @Microsoft Execution Containers (MXC), giving users a more secure way to work with agents!
Developers and IT administrators define the resources an agent can use and MXC uses the appropriate container to enforce those policies at runtime.
Read more about the policy-driven containment here blogs.windows.com/windowsdev…
Simular retweeted
Finally trying out robosecretary too.
Gave @sai_borg a task: learn about me through my public handles (I gave my github and X username), understand what I do, and build a client acquisition playbook around it.
Planning on doing cold emails going forward using Sai.
Will update if it works great.
If the outreach is personalized to your profile and actually relevant, there’s really no harm in automating the cold mails, right?
Simular retweeted
Three years into making computer use agents, I kept asking myself:
Why hasn't CUA reached super intelligence yet?
@sai_borg (Sai), which happens to sound like SI, has invented a new way to build computer use agents: not just more reliable, but also 99% cheaper
The answer lies in how humans function: build muscle memory to optimize for efficient output over time
We released a paper today on this new paradigm, called neuro-symbolic
Give it a read here: simular.ai/articles/real-wor…
Simular retweeted
We’re hosting an event about “New era of Personal Agent” at SF @Techweek_
partner with SEAMATE X @CoreSpeedHQ
And we invited：
• @Richard_simular — @SimularAI
• @ZongZiWang — @harvey
• @haosu_cs — @CoreSpeedHQ
• @samuel_spitz — @Replit
📍 Downtown SF
🗓 Oct 8 · 5–8 PM PT
See you there.
luma.com/seamate-tdd8
Simular retweeted
Agent demos fooled you 👎
It trips, it asks for help, and somehow you're the one copying, pasting and running the commands.
Meet your #Robosecretary. Meet your #SaiFleet.
Sai brings state-of-the-art computer use to your own computer - your apps, your OS, in a floating window that leaves your mouse alone.
Need more than one? Spin up to 100 cloud machines, each on the OS you pick, each on a different job. 100x the work, done without you.
Sai. The autonomous computer fleet at your command.
Free to start → sai.simular.ai
Simular retweeted
We told you what Sai can do. Now watch this.
Live on Product Hunt today — don't forget to UPVOTE and leave a COMMENT ❤️
🆙 producthunt.com/products/sim…
Simular retweeted
Cost / performance ratio matter; and it will matter more.
Exciting progress by @SimularAI
Our computer-use agent @sai_borg just beat Opus 5 and GPT-5.6 Sol on OSWorld 2.0.
Sai scored 73% on the CUA benchmark where each complex task takes a skilled human over an hour to execute.
And Sai did it at about 2/3 the cost per task of Opus and GPT 💸
The unlock is our neurosymbolic framework that pairs the exploratory power of neural networks with the logic of symbolic code, resulting in reliability and cost saving.
Read the full report: simular.ai/articles/sai-tops…
Simular retweeted
One pain point of computer-use agents is that they’re simply too expensive to run at scale.
Better performance matters, but so does the cost of getting there.
On OSWorld 2.0, Sai achieved the highest score among the compared models significantly lower cost than other frontier models.
For computer-use agents to move from impressive demos to everyday production, they need to be not just smart, but affordable enough to scale.
Intelligence per dollar matters.
Try us at Sai.work
Simular retweeted
Last December, our researchers built the first agent to reach human-level performance on OSWorld - Agent S, 72.6% against a human baseline of 72.36%
Today they've put us back at #1 on OSWorld 2, at less than two-thirds the cost per task
Kudos to the team! Let’s continue making AI affordable for everyone ✨
Our computer-use agent @sai_borg just beat Opus 5 and GPT-5.6 Sol on OSWorld 2.0.
Sai scored 73% on the CUA benchmark where each complex task takes a skilled human over an hour to execute.
And Sai did it at about 2/3 the cost per task of Opus and GPT 💸
The unlock is our neurosymbolic framework that pairs the exploratory power of neural networks with the logic of symbolic code, resulting in reliability and cost saving.
Read the full report: simular.ai/articles/sai-tops…
Simular retweeted
It's happening -- higher accuracy at two-thirds the cost. Reliability and cost-efficiency, both landing on Sai at the same time.
Full report below ⬇️
(Also: every time my wifi dropped and I ended up playing Chrome Dino, I kept thinking this game was *made* for Sai to play ;)
Our computer-use agent @sai_borg just beat Opus 5 and GPT-5.6 Sol on OSWorld 2.0.
Sai scored 73% on the CUA benchmark where each complex task takes a skilled human over an hour to execute.
And Sai did it at about 2/3 the cost per task of Opus and GPT 💸
The unlock is our neurosymbolic framework that pairs the exploratory power of neural networks with the logic of symbolic code, resulting in reliability and cost saving.
Read the full report: simular.ai/articles/sai-tops…
Simular retweeted
Thanks for the coverage @thenewstack
“Sai’s optimal cost-outcome tradeoff on OSWorld 2.0 is one step toward a future where we don’t need to settle for exorbitant prices to get the job done.” -- Jiachen Yang
thenewstack.io/sai-agent-osw…
