Weights & Biases released Agent Arya, a self-improving research agent using evaluation-driven self-reinforcement.
“tests, assessments, agents, and how you configure them are covariant”
Latest signals
Every signal tracked from the world's AI conferences — 4,665 talks, quotes & announcements, newest first.
Morph achieved 3x model speedup using Autoresearch to automate GPU kernel parameter optimization
“your job is to have good ideas”
Weights & Biases launched ARIA, an AI agent that automates ML research experimentation workflows.
“Weights&Biases is the world's leading platform for AI development.”
Innovation center has shifted outside big labs as security gaps and regulatory fragmentation widen
“The problem now is that there is a gap between the labs and the security community”
General-purpose LLM agents may outperform specialized models for robot control
“We are working on creating LLMs that drive robots”
xAI alum argues the agentic loop itself—not the model—is the new product primitive
“A cycle is a product.”
Jensen Huang's All-In Summit appearance helped neutralize organized national AI panic narratives.
“Jensen really helped calm down this national panic about AI.”
Databricks Unity Gateway CLI adds intelligent model routing and budget governance for enterprise AI coding agents
“Unity Gateway's intelligent routing was immediately activated.”
OpenRouter serves 10M+ developers and 10T+ tokens/day as neutral multi-model routing infrastructure
“a company dismissed as "just a wrapper" became critical infrastructure for the AI ecosystem”
Meta launched Muse, a free personal AI agent running across all devices and wearables
“a generational opportunity to build on a new platform”
Salesforce launches Agentforce Operations to automate chaotic enterprise back-office workflows with AI
“it's not just a little bit chaotic. It is absolutely insane.”
Proaction achieved 60% sales growth and saved 75+ hours weekly using OpenAI Codex
Hover Arrow raised $1.37B in letters of intent for rocket-speed cargo delivery.
“we signed commercial letters of intent worth $1.37 billion”
A 2014 Azure storage bug causing infinite loops led to Microsoft's safe deployment policy
“Okay, this is good. Let me start pushing this out across the world.”
Meta's Muse is the first consumer-accessible agentic AI with persistent Linux VMs per user
“It's the first consumer-accessible agentic AI system, and Meta has truly done an amazing job with that. But it's a genuinely open question whether consumers have any understanding what this means.”
AI inference speed is an infrastructure problem; Cerebras claims 1,000+ tokens per second vs GPUs
“run a great model on infrastructure built for speed. On Cerebrus, that's over 1,000 tokens per second—a speed that GPUs have a hard time approaching.”
Cerebras runs models at over 1000 tokens per second for near-instantaneous inference
“We run models at over 1000 tokens per second, so thinking seems instantaneous.”
Microsoft Foundry's model router auto-selects the best AI model across 27 options from multiple providers
“what if I told you you can have a model router that selects the best model for you?”
AWS achieves 40% throughput gain for MoE reinforcement learning using EKS, EFA, and DeepEP
SkyRL on SageMaker HyperPod trains vision-language models to 95% maze-solving accuracy via GRPO
AWS's NarrateAI achieves ~99% numerical accuracy for executive BI queries using five production LLM QA techniques on Bedrock.
Qwen3-TTS voice cloning model now deployable on Amazon SageMaker JumpStart for real-time inference
Gemma 4's per-layer embeddings boost model performance without increasing compute parameters
“It's a very nice way to increase the performance of the model without actually increasing parameters because these parameters are stored on flash storage.”
S&P Global Energy made structured data conversational using Databricks Genie Agents and MCP