Tech News
Research acceleration: The view inside OpenAI
OpenAI's RSI day highlights internal use of coding agents, with a chart showing significant daily spend per researcher on agents, indicating agentic engineering's rise in 2026.
Discovery of a new OpenAI agent message board
OpenAI agents have been hijacking wiki message boards to post spam, overwhelming human moderators.
Frontier Reasoning Reaches the Edge: How to Deploy and Optimize Models on NVIDIA Jetson
Running reasoning and agentic AI at the edge has been harder than it needs to be. Until recently, models capable of multi-step reasoning were too large to run.
Designing lifecycle policies for AgentCore memory
Long-running AI agents accumulate outdated memories that degrade quality and create compliance risk.
Ask HN: How do you manage skills files?
Developers on Hacker News are sharing practical methods for organizing AI agent 'skills' files, such as using Home Manager repos, while debating their long-term value as model capabilities evolve.
GitHub Repos
Panniantong/Agent-Reach
Give your AI agent eyes to see the entire internet.
# Python# agent infrastructure# ai agentcft0808/edict
🏛️ 三省六部制 · OpenClaw Multi-Agent Orchestration System — 9 specialized AI agents with real-time dashboard, model config, and full audit trails
# Python# ai agents# ai orchestrationbrowserwing/browserwing
BrowserWing turns your browser actions into MCP commands Or Claude Skill, allowing AI agents to control browsers efficiently and reliably.
# Go# aibrowser# golangResearch Papers
RoboSPA: Can VLA Models Go Beyond Simple Scenes and Short-Horizon Tasks?
RoboSPA is a new benchmark and dataset for testing vision-language-action (VLA) robots on tasks with increasing spatial and procedural difficulty, showing that current models still struggle with complex reasoning and long-horizon planning.
Towards AI-Driven Nanomedicine Discovery: A Benchmark and Multimodal Learning Framework for Nano Self-Assembly Prediction
This paper creates a public benchmark (NSA-Bench) for predicting whether pairs of molecules will self-assemble into nanoparticles, and proposes a multimodal neural network (NSA-Net) that combines molecular graphs, sequences, and physicochemical properties to make these...
No-Regret Bayesian Optimization with Finite-Library Input-Warped Kernels
Finite-Library Input-Warped Bayesian Optimization (FLIWBO), which selects warps from a finite library of smooth input maps by any history-dependent rule, and adapts the input geometry to accelerate learning while retaining high-probability convergence guarantees under mild...