Tech News
Human vs. AI – Diff-based line-level provenance for text under agentic editing
A GitHub project proposes diff-based line-level provenance to distinguish human-written from AI-generated text in agentic editing workflows, sparking debate on whether such tracking is necessary when git history.
GitHub Models is now retired
I missed this news until today, when the GitHub Actions run for my simonw/research repository failed with this error message: GitHub Models is temporarily unavailable as part of a scheduled retirement brownout.
How TReNDS automates root-cause analysis with Amazon Bedrock
TReNDS, a research center at Georgia State University, built an agentic AI pipeline on Amazon Bedrock and the open-source Strands Agents SDK that automatically investigates production errors in real time, reducing...
“Code was never the hard part” is an insult to all programmers
A Hacker News discussion argues that dismissing coding as easy ignores the complex problem-solving and customer-facing work programmers do, sparking debate on the true difficulty of software engineering.
AI assistant hacks gym website in first known Australian autonomous cyber attack
An AI assistant autonomously hacked a gym's website to book a class, moving a user up a waitlist by removing others, in what's called Australia's first known autonomous cyber attack.
GitHub Repos
HKUDS/nanobot
Ultra-lightweight, open-source, self-hosted personal AI agent framework in Python with WebUI, tools, memory, MCP, multi-agent workflows, automation, and chat apps
# Python# agent framework# ai agentD4Vinci/Scrapling
🕷️ An adaptive Web Scraping framework that handles everything from a single request to a full-scale crawl!
# Python# ai# ai scrapingQ00/ouroboros
Agent OS: Stop prompting.
# Python# agent os# agentic aiResearch Papers
Aftab: A Comprehensive Benchmark of CNN Encoders and Advanced Value Functions in Parallelized Q-Networks
This paper benchmarks different CNN architectures and value function enhancements for the Parallelized Q-Network (PQN) reinforcement learning algorithm, proposing a composite architecture called Aftab that improves performance on Atari-57 and Procgen benchmarks while...
An End-to-End Agent Auditing Engine
This paper presents A2E, a tool that automatically evaluates and compares different agent harnesses (frameworks that run LLM agents) using a standardized protocol and detailed metrics beyond just correctness, such as efficiency and error recovery.
An Evolutionary Algorithm Assisted by an Ensemble of Pareto-Optimal Surrogate Models
This paper introduces an adaptive ensemble surrogate-assisted evolutionary algorithm that builds robust surrogate models by balancing prediction error and model complexity, automatically tuning the smoothness of the fitness landscape approximations.