The Android Open Harness Project (AOHP) is an open-source operating system-level agent harness built on the Android Open Source Project. It addresses the mismatch between current application-centric operating systems and the needs of autonomous AI agents by treating agents as first-class OS actors. The design introduces three key mechanisms: personalized service composition, efficient agent interfaces, and secure information flow. These features enable adaptive user interfaces and agent-friendly runtime environments while preserving the existing Android ecosystem. Preliminary experiments on challenging tasks demonstrate significant performance improvements over conventional systems. Specifically, AOHP achieved a 21.12% increase in task completion rates compared to baseline methods. It also reduced token execution costs by 51.55%, highlighting its efficiency gains. Furthermore, the system showed improved compliance with security policies during agent-mediated interactions.
AOHP: An Open-Source OS-Level Agent Harness for Personalized, Efficient and Secure Interaction
HealthClaw: open-source agent with self-evolving memory for longitudinal personal health management
Researchers developed HealthClaw, an open-source agent architecture designed to update support as a person's routines, preferences, measurements, and risks change over time. It separates shared safety rules and medical knowledge from private longitudinal memory containing profile facts, reusable procedures, and episodic traces.
Tmax: A Simple RL Recipe for Terminal Agents
Tmax presents the strongest open RL recipe for terminal agents, achieving 27% on Terminal-Bench 2.0 with only 9B parameters. It uses a novel data taxonomy to generate over 2.5x more terminal environments than prior datasets, enabling efficient training with a simple, outcome-only recipe. The dataset, models, and code are open-sourced at https://github.com/hamishivi/tmax.
Technical Taxonomy of LLM Agent Communication Protocols
A new taxonomy classifies LLM agent communication protocols across five dimensions: counterparty, payload, interaction state, discovery mechanism, and schema flexibility. Analysis shows hybrid payloads, session-state persistence, and runtime schema negotiation are common, with decentralized discovery remaining rare. The study predicts short-term convergence toward unified agent-to-agent and agent-to-context protocols, and long-term evolution toward a federated, layered protocol stack.
OpenAI claims AI-assisted Navier-Stokes singularity result using 10,000 agents
OpenAI-affiliated accounts reported that an AI-assisted effort produced a result related to the Navier-Stokes existence and regularity problem, one of the Millennium Prize Problems. The claim states that approximately 10,000 agents collaborated on the task after being trained for roughly a year using multi-agent reinforcement learning.
OpenAI internal system proves Navier–Stokes equations can develop finite-time singularities
OpenAI reports that an internal AI system has produced a proof showing that the dynamics of the Navier-Stokes equations for fluid motion can develop a singularity in finite time. The solution, derived by a multi-agent system powered by a model significantly more capable than GPT-6 Astra, establishes statement "C" of the Millennium Prize formulation.