Speko launched on July 29th as a benchmark-based router for voice AI models, giving developers a single gateway that selects speech and language models using continuously updated benchmarks. Founder B…
AntGroup released Ling 3.0 Flash, a 124-billion-parameter hybrid-reasoning Mixture-of-Experts model with 5.1 billion active parameters per token, matching or beating its 1-trillion-parameter flagship …
Agent Detective, a new open-source evaluation framework for multi-agent systems, identifies the first node where quality breaks in a pipeline by ingesting standard OpenTelemetry traces and rebuilding …
Together AI announced an $800 million Series C led by Aramco Ventures, with Nvidia, Vista Equity Partners, General Catalyst, Emergence Capital, and Salesforce Ventures participating, valuing the compa…
Aris Alexis Giachnis is building Echologue, an AI voice journal that stores its permanent archive locally on the user's device while routing consented transcription and AI tasks to cloud providers inc…
Ricky Rojas released The Groundhog Trap, an AI governance framework prototype (v0.1) that routes prompts through multiple independent large language models to generate auditable consensus scores and d…
China's AI advantage lies in affordability and accessibility rather than peak performance, with Chinese models processing 18 trillion tokens weekly on OpenRouter by June, triple the U.S. total, accord…
A developer building Hitou, a personalized song generation service, documented real-world failure modes of AI APIs that reached production. The team found that LLM API responses with status 200 and fi…
LLM inference profitability depends on the trade-off between batch size and GPU count, which determines token latency and cost per million tokens. Applying this model to Kimi K3, which requires at lea…
Moonshot AI released the model weights for Kimi K3, a 2.8 trillion parameter language model, under a custom licence on July 27, 2026, alongside a 47-page technical report. The Beijing-based startup's …
Y Combinator-backed startup raises $13.5 million to help enterprises control LLM token spending as the tokenmaxxing trend collapses under rising costs. Uber blew through its 2026 AI coding budget by A…
Moonshot AI released the open weights for Kimi K3 on July 27, a 2.8 trillion parameter Mixture-of-Experts model with 50 billion active parameters per forward pass, requiring a minimum of 8x H100 80GB …
A developer released MoilStack .md, an open-source Markdown editor with inline AI editing that supports local models via Ollama and cloud APIs from OpenAI, Anthropic, Google, and others. The editor st…
Answerwatch, a new open-source tool, lets users track changes in what AI models recommend by running the same prompt across models like OpenAI, Anthropic, and Google, storing results locally in SQLite…
The United States could face "massive capital destruction" from its artificial intelligence investment boom as Chinese open-source AI models increasingly challenge US hyperscalers, according to a Jeff…
Cursor shipped Moonshot AI's Kimi K3, a 2.8 trillion-parameter mixture-of-experts model with a 1 million-token context window and native vision, on its launch day July 27, making the open-weight model…
Moonshot AI released Kimi-K3 with a new license that requires large Model-as-a-Service businesses (those exceeding $20 million in aggregate revenue over any consecutive 12 months) to enter a separate …
60 Minute Creator launches a node-based AI content workflow builder that lets users create reusable workflows with their own API keys from providers like fal.ai, Kie AI, and OpenRouter. The tool offer…
A new bipartisan U.S. congressional bill requiring AI companies to install a kill switch for advanced models would be powerless against open-weight AI models, which can be deployed locally and modifie…
Claude 3.5 Sonnet is currently the hardest large language model to jailbreak due to its strong attention mechanism and resistance to persona-shift attacks, according to a Machine Learning Forum discus…