Sai✓

Simular @SimularAI

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:

Prompt Engineer 48 @prompt48

An AI agent that writes its own tools beat clicking cell by cell. I tested the idea behind Agent S4 on my 8GB laptop, and it also crashed on one sheet. Full breakdown 👇

DAIR.AI @dair_ai

Guitar Cat + LLM 「Will @ SGLang Summit in SF」 @GenAI_is_real

The interface matters. Give your agent a computer. It stops block me when doing compute use 😂 Connected Codex to Simular’s Sai MCP, sent a plain-English task, and watched Sai operate Windows Calculator on a hosted desktop. It entered the calculations through the app and left both results visible in History. For developers, having the computer and desktop control available behind MCP is pretty useful. I can hand my agent a desktop task without building and maintaining the VM, screenshot pipeline, and action loop myself. Simular’s Agent S4 research also caught my attention: keep the model and agent harness fixed, then improve the interfaces the agent works through. Alongside clicks and typing, better tools can turn repetitive actions into a single operation. I think that direction has a lot of potential for vertical agents too. This video is a small, real Sai product demo; it doesn’t reproduce the S4 experiments. Waiting was cut from the recording. Separately, Simular reports the highest partial score on OSWorld 2.1—79.38%—for Sai with Claude Opus 5, at $14.34 per benchmark task. Those are vendor benchmark figures, separate from this demo.

Simular @SimularAI

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:

Soham @soham_nayak04

I’m building SohamLaunches solo. Every time I ship a change, I have to click through every page myself to check nothing broke. It’s boring, so I skip it, and that’s how bugs reach users. So I gave @sai_borg a computer and told it to QA my site. One prompt through Claude Code, and it opened the site on a Windows cloud computer and clicked through every page. Live view in the video. 9 pages checked, 106 directories scrolled, 5 minutes, $0.03. It flagged one 404 on /badges. Turned out to be my wrong guess, the route only exists per product. One false alarm, nothing actually broken. It also stopped at the login wall without trying to get through. Good. Would I use it for real QA? Yes, for the boring “click everything and tell me what breaks” pass. I’d still review what it flags.