Tech News
GPT-5.6 Sol Ultra will be in Codex
OpenAI introduces GPT-5.6 Sol Ultra with a new ultra mode using subagents for complex tasks, alongside reports of halved inference costs.
huggingface/transformers Release v5.13.0
Release v5.13.0 New Model additions KimiK 2.5, 2.6, and 2.7 This release includes the architecture for Kimi 2.5 which is used by 2.5-2.7: Kimi K2.5 is an open-source, native multimodal agentic model that advances prac.
sqlite-utils 4.0rc2, mostly written by Claude Fable (for about $149.25)
I wrote about the sqlite-utils 4.0rc1 release a couple of weeks ago. Since we only have Claude Fable on our Max subscriptions for a few more days, I decided to see if it could help me get to a 4.0 stable release.
EU Council forces Chat Control via fast-track
The EU Council is fast-tracking Chat Control 1.0, allowing messaging providers to scan chats for harmful content, raising privacy concerns.
New AI tutor achieves 0.71-1.30 SD effect size in Dartmouth course [pdf]
An AI tutor shows 0.71-1.30 SD effect size in a Dartmouth course, but skepticism arises due to small sample size and lack of randomization.
GitHub Repos
koala73/worldmonitor
Real-time global intelligence dashboard.
# TypeScript# agent# aidiegosouzapw/OmniRoute
Never stop coding.
# TypeScript# a2a# ai agentsmcp-use/mcp-use
The fullstack MCP framework to develop MCP Apps for ChatGPT / Claude & MCP Servers for AI Agents.
# TypeScript# agentic framework# aiResearch Papers
ReContext: Recursive Evidence Replay as LLM Harness for Long-Context Reasoning
ReContext improves LLM reasoning over long contexts by recursively replaying relevant evidence from the input without training or external memory, boosting performance on 128K-length tasks.
SPLC: Social Preference Learning for Crowd Robot Navigation
This paper proposes a method to automatically learn social preferences for robot navigation in crowds, avoiding manual reward design, and shows improvements in simulations and real-world tests with TurtleBot4.
DL-VINS-Factory: A Modular Framework for Learned Visual Front-Ends in Visual-Inertial SLAM
DL-VINS-Factory is presented, a unified framework that integrates learned feature extractors with either Lucas--Kanade optical-flow tracking or LightGlue descriptor matching, and results show that learned front-ends are viable for real-time embedded VI-SLAM, but are not...