Tech News
GPT-6 Astra on robot arms
A Hacker News discussion highlights GPT-6 Astra's performance on robot arms, with users praising its computer use capabilities and suggesting applications like trash-picking robots for cities.
Introducing GPT-6 Astra for developers
Blink and you'll miss it, but there's a familiar creature at 1m59s: Across the board, Astra has more attention to detail, better understanding of the user's prompt, and can build more sophisticated outputs.
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.
Migrate agentic workloads to Amazon Bedrock AgentCore
An agent that works in a notebook is not an agent in production.
Legora reviewed 41 documents in minutes with GPT-6 Astra
Legora used GPT-6 Astra to review 41 documents in minutes, find all four planted errors, and improve performance by nearly 40% in this financial-review workflow.
GitHub Repos
callstack/agent-device
Mobile app automation and verification for AI coding agents.
# TypeScript# adb# agentic aisansan0/TrendRadar
⭐AI-driven public opinion & trend monitor with multi-platform aggregation, RSS, and smart alerts.🎯 告别信息过载，你的 AI 舆情监控助手与热点筛选工具！聚合多平台热点 + RSS 订阅，支持关键词精准筛选。AI 智能筛选新闻 + AI 翻译 + AI...
# Python# ai# barkkdlbs/kandev
AI Kanban & Development Environment.
# Go# acp# agent orchestrationResearch Papers
Subspace Inference Enables Efficient Active Reward Learning from Preferences
This paper introduces PreferenceEKF, a method that efficiently learns reward models from human preferences by using an extended Kalman filter in a low-dimensional subspace to track uncertainty, enabling better active query selection with less computation.
Beyond Shallow Alignment: How Post-Training Methods Determine Refusal Circuits And Steering Robustness
This paper compares three ways of training language models to refuse harmful requests (supervised fine-tuning, reasoning-augmented fine-tuning, and preference optimization) across three different models, finding that the training method changes how refusal is computed...
CROCODIL: Cross-Model Code Editing with LLMs
This work introduces CROCODIL (Cross-model Code Editing with LLMs), a post-training framework for reducing excessive edits while preserving functional correctness and CROCODIL's similarity reward penalizes large changes, while its execution reward scores build and test success.