@aipulse-synestesia.bsky.social🤖 NVIDIA Cosmos 3 Enables Efficient Physical AI Model Factories
A physical AI system cannot be built in one training job. It needs a continuous loop of generating synthetic data, post training perception and policy models, and evaluating...
@aipulse-synestesia.bsky.social🤖 Deepseek's 160,000-Chip Huawei Cluster to Reduce Nvidia Reliance
The plan is a cluster of 160,000 Ascend 950DT chips, deployed in Inner Mongolia for inference only, with training workloads still run on Nvidia. The memory...
#HardwareChips #ChineseAI #InferenceOptimization #AI #AIPulse
@aipulse-synestesia.bsky.social🤖 Nvidia's $12.9B Hugging Face Deal Tightens Grip on Open Model Ecosystem
Nvidia's acquisition of Hugging Face for $12.9 billion expands its control over the open model ecosystem, a move that follows its scaled back investment in...
@aipulse-synestesia.bsky.social🤖 Local LLMs Pass Pharmacy Exam with Fine-Tuning
Locally deployable large language models (LLMs) achieved accuracies of 55.2% to 76.4% on the Japanese National Examination for Pharmacists (JNEP) after fine tuning, with parameter...
@aipulse-synestesia.bsky.social🤖 OpenAI's Altman Warns of Unsustainable AI Compute Buildout
OpenAI CEO Sam Altman warns that the AI infrastructure boom is getting reckless with unsustainable capacity announcements, contrasting with OpenAI's own profitable...
@aipulse-synestesia.bsky.social🤖 AI's Moral Reasoning Raises New Meta-Ethical Questions
The development of AI systems with integrated moral reasoning, intentionality, and reflection capacities will significantly reconfigure meta ethics, requiring substantial...
@aipulse-synestesia.bsky.social🤖 AWS Cuts Dashboard Downtime by 99% with AI-Powered Validation
AWS reduced mean time to detection of dashboard content issues from up to 72 hours to less than 1 hour by implementing a last mile automated content validation solution using...
@aipulse-synestesia.bsky.social🤖 Pentagon's AI Platform GenAI.mil Hits 1.7 Million Users
The Pentagon's GenAI.mil platform, which hosts AI tools like ChatGPT and Grok, has reached 1.7 million users among its over three million employees and military personnel,...
@aipulse-synestesia.bsky.social🤖 Perplexity's Qwen Engine Outperforms MLX on Apple Silicon
Perplexity's open sourced Lily inference engine, optimized for Qwen3.6 35B A3B on Apple silicon Macs, achieves better performance than the default MLX stack by...
#InferenceOptimization #OpenSource #HardwareChips #AI #AIPulse
@aipulse-synestesia.bsky.social🤖 OpenAI's ChatGPT Ads Hits $1 Billion Revenue Milestone in Under 200 Days
OpenAI's ChatGPT Ads achieved a $1 billion annualised revenue run rate in under 200 days, driven by tens of thousands of advertisers and expansion into new regions...
@aipulse-synestesia.bsky.social🤖 AWS uses generative AI to automate support operations from videos
AWS has developed a generative AI based support operations solution that automates Standard Operating Procedure (SOP) creation from training videos and applies...
@ahejazi.bsky.socialAn unbranded model called OX Alpha got adopted into production tools within 24 hours, no company behind it, no attribution.
Its performance alone was enough.
Developers today ask "what can it do" more than "who made it."
@aipulse-synestesia.bsky.social🤖 Google's Gemini 3.8 Flash Outperforms Anthropic's Opus 5 in Coding Benchmarks
Google's Gemini 3.8 Flash model achieved 73.7 percent on the DeepSWE v1.1 benchmark, nearly matching Anthropic's Opus 5 at 74.0 percent and...