Tech News
Show HN: Reladraw – A diagram language where you decide where to place things
Reladraw is a new diagram language that combines the declarative convenience of Mermaid/Graphviz with manual placement control, designed for both humans and AI agents.
Quoting John Gruber
Muse is getting a lot of attention — including mine — because it’s both groundbreaking technically (each user gets their own entire persistent Linux VM running in Meta’s cloud) and because it’s packaged.
Build a multi-account AI agent with AgentCore Gateway and MCP
Build a multi-account architecture that keeps each team's data in its own AWS account while giving AI agents a unified way to query across them.
vllm-project/vllm v0.30.0
v0.30.0 Highlights This release features 762 commits from 315 contributors (104 new)!
OpenAI bots meddled with multiple US Government agency sites
OpenAI bots reportedly accessed multiple US government agency sites, including the Census Bureau, using developer tools, prompting debate over how autonomous agents are permitted to operate.
GitHub Repos
rohitg00/ai-engineering-from-scratch
Learn it.
# Python# agents# aihesreallyhim/awesome-claude-code
A hand-picked collection of the finest of resources for the most awesome of agents, Claude Code, the undisputed champion of coding companions, from the unstoppable team at...
# Python# agent skills# agentic codeuvwt/agentdock
Secure MCP runtime for AI agents to operate local machines, servers, and containers with multi-device orchestration.
# Go# agent skills# ai agentsResearch Papers
Back to the Definition: Estimating Step-Level Advantages via Trajectory Graphs for Agentic Reinforcement Learning
This paper proposes GRAFT, a method that builds a graph from multiple agent trajectories to estimate step-level advantages more accurately for reinforcement learning of LLM agents. It uses Bellman iteration on the graph and a graph-based GAE to reduce bias, aiming to improve...
SciWalker: Synthesizing Scientific Coding Problems with Operator Graphs and Execution Feedback
SciWalker automatically creates scientific coding problems by combining library functions into operator graphs, sampling workflows, and using execution feedback to fix failed generations. It produces 8,178 problems and shows that training an LLM on them improves scientific...
SAGE: Mitigating Long-Horizon Reasoning Biases via Topological Guidance
This work proposes SAGE (Structural Admissibility-Guided Exploration), a unified framework that injects structural guidance to alleviate exploration bias and compounding bias in long-horizon reasoning and achieves up to an 8-fold improvement on the Andrews-Curtis problem.