Daniel Zhang

Agents Research @Scale_AI

[Scale Labs @ScaleAILabs]

📣Call for contributions + co-authorship!

RSI Bench is our ongoing effort to evaluate whether AI agents can develop the capabilities required to advance AI R&D.

Got a frontier AI research problem on your mind? Submit it to RSI Bench, and set the standard the whole industry uses to measure progress!

Contributors receive:
→ Co-authorship on the RSI Bench research paper
→ $2,000 per accepted task for the initial 50 tasks
→ Modal compute credits to build + iterate
→ Access to the RSI Bench research community

Register below for full requirements.

[Daniel Zhang @danielyuez]

@Yminglai @RazvanDuu @liuying04 @utkarsh4430 @XiY246 🐮

[Chetan Rane @the_chet2an]

Who’s allocating budget for human data?
2019: AV builders
2022+: big 7 labs
2025: ~100 neo labs, physical AI
2026: every enterprise

dying market though. nothing long term worth building here

[Minglai Yang @Yminglai]

RL against rubrics quietly reward-hacks: the training judge keeps scoring higher while true quality falls.

One-line fix: randomly drop part of the rubric each step, so the policy never optimize the same set of rubrics twice.

📄 https://t.co/BSrzyJOciF

[Yangqing Jia @jiayq]

We are starting Intent Lab, building an autonomous team we call "fleet" that turns intent into production software. Today we are sharing some early results: the fastest GLM5.2 inference engine, one shot database creation, and a fully verified agent filesystem.

https://t.co/CtwtypE0rH

[Aakash Sabharwal @aakashsabharwal]

Proud to support this on behalf of @scale_AI

Open models are critical to American AI leadership and to building reliable AI.

Great to join so many across the industry in signing on.

[Scale Labs @ScaleAILabs]

Frontier-Bench (formerly Terminal-Bench) by @harborframework, an open-source benchmark that stress-tests AI agents on real command-line tasks, is now live.

We're proud to be the top contributor to the launch set with tasks designed to pinpoint exactly where an agent breaks.

[Chetan Rane @the_chet2an]

Lots of recent discussion on MCP vs CLI as the better tool use interface. We ran them head to head:

- CLI helps weak models with trial-and-refine loops
- Search & retrieval in larger environments is more reliable with CLI
- MCP wins on precise reads and writes
- CLI is generally much costlier

As models improve, the gap converges

https://t.co/R1Goj46b0o

[Aakash Sabharwal @aakashsabharwal]

SWE Atlas from @ScaleAILabs features in GLM-5.2 launch!

[Scale Labs @ScaleAILabs]

We're back from Seoul!

Great sharing our research on agent training, self-optimization, and evaluation through our paper presentations, lightning talks, and reception at @ICMLConf.

The momentum toward bringing AI agents into real professional domains is massive, and we're thrilled to drive that conversation.

Next stop: @NeurIPSConf 🇳🇿

[Scale Labs @ScaleAILabs]

We appreciate the community's feedback on SWE-Bench Pro.

Much of it maps to changes already underway in v1.1, which we've been building for a while. Keeping evals current with frontier models is hard, and we're always iterating.

SWE-Bench Pro Verified coming soon. 👀