PhAI Labs Released ScienceBuddy AI Workspace
The research firm launched a browser-based platform designed to integrate scientific data and analysis tools.
Updated on Sept. 24, 2026 in Artificial Intelligence
Live Poll
Do you trust AI research tools that actively learn from user feedback?
PhAI Labs has released the ScienceBuddy Preview, a new browser-based workspace that integrates scientific papers, data, and analytical tools. Available in English and Chinese, the platform is designed to improve AI models through direct researcher collaboration.
Why it matters
The system aims to enhance AI performance by utilizing researcher feedback and iterative learning loops. By incorporating these collaborative elements, the platform seeks to refine task models and software harnesses effectively.
The workspace features 224 tools organized into 22 functional modules. Evaluation of the system showed single-attempt accuracy improved from 42.2% to 73.3% across 895 scientific tasks.
The players
PhAI Labs
PhAI Labs is a research organization headquartered in Palo Alto, California, that focuses on developing advanced artificial intelligence systems.
The details
The platform functions using two distinct learning loops that refine both the task model and the software harness based on researcher session data. Research code and the corresponding technical report are available on GitHub and arXiv under the MIT License.
Timeline
September 24, 2026: PhAI Labs released the ScienceBuddy Preview workspace.
The Tech Race
This release follows the standard industry pattern of using the arXiv preprint server for technical validation before broader tool deployment. It represents a shift toward specialized AI environments that replace general-purpose chatbots for complex academic workflows.
Researchers can now utilize the platform's 224 integrated tools to streamline their analysis and data processing tasks in English or Chinese. The adoption of this interface may improve workflow efficiency for users currently reliant on fragmented software solutions.
The takeaway
The move toward specialized scientific AI workspaces highlights an industry focus on integrating human feedback directly into learning loops. Users looking to improve their research accuracy may benefit from adopting platforms that provide structured, data-driven software harnesses.
Further reading
For more on the current state of industry research, visit the Artificial Intelligence section.
More information
Researchers and developers can access the full ScienceBuddy workspace access to begin testing the platform.
Live Poll
Do you trust AI research tools that actively learn from user feedback?
