Google Research’s Harness² introduces a new approach to AI agent optimization by allowing agents to refine their own prompts, tools, memory and orchestration logic without modifying model weights. Early benchmarks show promising performance improvements, but the open-source framework also raises questions about security, API costs and governance as enterprises explore self-improving AI agents.
NVIDIA today released the full Run:ai platform as open-source software. […]
Microsoft Decision-1 challenges TypeSafe’s Jev with fast, low-cost AI decision scoring for agent routing, classification and verification. Explore the benefits and risks.
Every AI agent that does real work needs real access. […]
Ripple has added Machine Payments Protocol support to its XRP Ledger AI Starter Kit, giving AI agents a standards-based way to pay for services using XRP or RLUSD without human checkout flows.
NVIDIA’s cuda-oxide and cutile-rs bring native Rust to GPU kernels, combining CUDA-level performance with compile-time memory safety for AI infrastructure.
Microsoft Agent Framework Channels lets developers expose one AI agent across Telegram, OpenAI Responses, A2A and MCP without rebuilding each integration.
Cursor Router automatically matches coding tasks with the most suitable AI model, helping engineering teams cut costs without sacrificing quality.
Celeris, a new AI research lab backed by Lightspeed Venture […]
Anthropic cut more than 80% of Claude Code’s system prompt, signaling a shift from rigid prompt rules to leaner, more adaptive context engineering.
