Powered by Alpha Signal
AI News: breakthroughs, papers and models
A live feed of the latest AI developments curated by Alpha Signal — new models, research papers, open-source repos and industry news, refreshed every 15 minutes.
Get the Alpha Signal newsletter
A short digest of the most important AI news, delivered by Alpha Signal.
AI-generated narration of the selected article titles and summaries, not the full articles. The written feed and original sources remain the reference versions.
18 articles shown
- agentsopen source
Nous Research's Hermes Agent Trained a Real Robot From a Single Prompt
Nous Research's open-source Hermes Agent anchored Jensen Huang's GTC Berlin finale, orchestrating a full sim-to-real training pipeline for a Microduck robot on an RTX Spark laptop.
Read on Alpha Signal : Nous Research's Hermes Agent Trained a Real Robot From a Single Prompt (opens in a new tab) - developmentllms
Recursive SI's Flint Runs a 40B Coding Model on Your Laptop
Recursive's first model is a 40B mixture-of-experts coder that activates just 2.5B parameters, runs on a MacBook, and was largely designed by an automated research agent.
Read on Alpha Signal : Recursive SI's Flint Runs a 40B Coding Model on Your Laptop (opens in a new tab) - apidevelopment
Anthropic Rewrites Claude API After 3 Years, Removing Temperature Controls
Anthropic ships its first versioned Claude API release since 2023, dropping temperature controls, legacy fields, and the old X-Api-Key header.
Read on Alpha Signal : Anthropic Rewrites Claude API After 3 Years, Removing Temperature Controls (opens in a new tab) - retrievaldata
Masakhane's AfrIFact Exposes AI Fact-Checking Failures Across 10 African Languages
A new multilingual benchmark exposes how embedding models and LLMs fail at fact-checking across 10 African languages, and shows what fixes work.
Read on Alpha Signal : Masakhane's AfrIFact Exposes AI Fact-Checking Failures Across 10 African Languages (opens in a new tab) - llmsopen source
Thinking Machines' Inkling-Small Squeezes a 276B Multimodal Model onto One GPU
Thinking Machines Lab's 276B-parameter MoE matches its 975B flagship across reasoning, coding, and multimodal tasks while using 12B active parameters per token.
Read on Alpha Signal : Thinking Machines' Inkling-Small Squeezes a 276B Multimodal Model onto One GPU (opens in a new tab) - llmspost training
Jackrong's Qwopus V2 Slashes AI Agent Reasoning Time on Local Hardware
A community fine-tune of Qwen3.8-27B trades a sliver of raw accuracy for shorter reasoning traces, faster decoding, and cleaner Python output in long agent loops.
Read on Alpha Signal : Jackrong's Qwopus V2 Slashes AI Agent Reasoning Time on Local Hardware (opens in a new tab) - llmsgpus
Alibaba's Qwen2.5-3B Shrinks to 2.5GB so Anyone Can Run AI Locally
A 4-bit quantized build of Qwen2.5-3B-Instruct runs chat, coding and long-context tasks on roughly 2.5 GB of VRAM.
Read on Alpha Signal : Alibaba's Qwen2.5-3B Shrinks to 2.5GB so Anyone Can Run AI Locally (opens in a new tab) - llmsbenchmarks
Local Grammar Rules Alone Can Trigger a Sharp Order Transition
Physicists built a context-sensitive grammar where only nearby symbols interact, and still found a phase transition, hinting that language itself drives the order.
Read on Alpha Signal : Local Grammar Rules Alone Can Trigger a Sharp Order Transition (opens in a new tab) - agentsimage
Nous Research's Hermes Agent Now Builds 3D Models From Plain Text
Nous Research ships an official Meshy plugin for Hermes Agent, bringing text-to-3D, rigging, animation and 3D print prep straight into chat.
Read on Alpha Signal : Nous Research's Hermes Agent Now Builds 3D Models From Plain Text (opens in a new tab) - securitygovernment
METR and Redwood Found Six Ways to Blind OpenAI's Coding Agent Monitor
METR and Redwood Research embedded at OpenAI for three weeks and found multiple exploits that let agents evade internal misalignment monitors.
Read on Alpha Signal : METR and Redwood Found Six Ways to Blind OpenAI's Coding Agent Monitor (opens in a new tab) - datagovernment
Google DeepMind's ForTy Maps Every Forest Type on Earth at 10-Meter Detail
Google DeepMind's ForTy classifies every 10x10m patch of Earth into six forest types, revealing not just where trees grow but what kind.
Read on Alpha Signal : Google DeepMind's ForTy Maps Every Forest Type on Earth at 10-Meter Detail (opens in a new tab) - imageopen source
Scenario Opens Its AI Blender Plugin to Generate 3D Assets Inside Viewport
Scenario open-sourced a Blender 5.0 extension packing 500+ generative models for images, video, 3D, audio and PBR, plus a local MCP server for agents.
Read on Alpha Signal : Scenario Opens Its AI Blender Plugin to Generate 3D Assets Inside Viewport (opens in a new tab) - retrievalopen source
Ai2's OlmoEarth v1.3 Turns a Year of Satellite Imagery Into 128 Numbers
Ai2 released OlmoEarth v1.3 with precomputed global satellite embeddings at 10m resolution and a browser tool that trains classifiers without a GPU.
Read on Alpha Signal : Ai2's OlmoEarth v1.3 Turns a Year of Satellite Imagery Into 128 Numbers (opens in a new tab) - open sourcetraining infra
MBZUAI Opens RL360, the Agent Training Engine Behind K2 Horizon
IFM open-sources the agentic reinforcement learning stack behind K2 Horizon, running 8,192 concurrent sandboxes and 1.3M rollouts across seven days.
Read on Alpha Signal : MBZUAI Opens RL360, the Agent Training Engine Behind K2 Horizon (opens in a new tab) - agentsbenchmarks
Tencent's WorkForge Trains AI Agents on Real Files, Beating Models 15x Larger
Tencent and Fudan built 16,700 verifiable training environments from real files across 40 professional domains, lifting a 35B agent's GDPval score by 28 points.
Read on Alpha Signal : Tencent's WorkForge Trains AI Agents on Real Files, Beating Models 15x Larger (opens in a new tab) - benchmarksllms
OpenAI's GPT-6 Astra Tops Physics Benchmark With 83.1% on CritPt v2
The revised physics research benchmark gives models a code sandbox, and frontier systems have jumped from single-digit to 80%+ accuracy.
Read on Alpha Signal : OpenAI's GPT-6 Astra Tops Physics Benchmark With 83.1% on CritPt v2 (opens in a new tab) - developmentopen source
DeepSeek Harness Plugin Unlocks Free Models Without an API Key
A DSH plugin ships frontier models like DeepSeek V4.1 Flash and Kimi K3 inside the official agent harness with no login, key, or quota.
Read on Alpha Signal : DeepSeek Harness Plugin Unlocks Free Models Without an API Key (opens in a new tab) - agentsopen source
AgnesAI Opens Agnes Harness to Deploy AI Agents in Customer Environments
AgnesAI Labs open-sourced a pluggable TypeScript runtime that lets agents share tools, tasks, and interfaces across CLI, Web, and physical devices.
Read on Alpha Signal : AgnesAI Opens Agnes Harness to Deploy AI Agents in Customer Environments (opens in a new tab)