Tech News
You said no MCP
A blog post and Hacker News discussion about reversing a prior rejection of MCP (Model Context Protocol), highlighting real-world uses where MCP lets local models like Qwen configure complex macOS apps.
huggingface/transformers Release 5.18.0
New Model additions Nemotron 3 Diarization Nemotron 3 Diarization is an open-weight streaming speaker diarization model designed to determine "who spoke when" in real-world audio.
Build a multi-agent music production pipeline on Amazon Bedrock AgentCore Runtime Instances
Amazon Bedrock AgentCore Runtime Instances gives multi-agent workflows AWS managed EC2 infrastructure with GPUs, persistent volumes, and multi-day sessions.
Tracing Agent Harness Behavior with NVIDIA NeMo Relay
An agent can finish a task and still take an inefficient path. A failed search can trigger another search. A truncated file read can lead to a command fetching.
Disrupting a coordinated model-distillation campaign
Learn how OpenAI disrupted a campaign to extract protected model reasoning and is strengthening defenses against adversarial distillation.
GitHub Repos
cobusgreyling/loop-engineering
Practical patterns, starters & CLI tools for loop engineering with AI coding agents.
# TypeScript# agentic ai# ai agentsMemPalace/mempalace
The best-benchmarked open-source AI memory system.
# Python# ai# chromadbusewhale/Whale
Whale — blazingly fast, terminal-first AI coding agent for DeepSeek.
# Go# coding agent# deepseekResearch Papers
AutoDataBench: A Data-centric Testbed for Accelerating Auto Research
This paper introduces AutoDataBench, a controlled benchmark for testing how well AI agents can improve training data rather than models. It checks whether agents can diagnose, organize, and construct better data, and whether they understand the effects of their changes.
NeuroDyn-EEG: An Interpretable Pre-trained Model for EEG Based on Neural Dynamics
This paper builds a pretrained model that turns standard scalp EEG into interpretable brain-dynamics parameters using a biophysical simulation model. It shows competitive disease classification and suggests region-specific changes in Alzheimer's and depression, but the work...
On the Relaxation of Conditional Independence Assumption for Image Segmentation
A new method makes image segmentation models better at matching the Dice/IoU scores used to grade them, by accounting for how nearby pixels' labels relate to each other instead of assuming they're independent.