Introducing Voice Isolator
Inworld AI released Voice Isolator, its first open-source model, under the Apache 2.0 license with code and weights on GitHub and Hugging Face. Voice Isolator runs at 48 kHz and ships as two checkpoints: an offline model…
Inworld AI released Voice Isolator, its first open-source model, under the Apache 2.0 license with code and weights on GitHub and Hugging Face. Voice Isolator runs at 48 kHz and ships as two checkpoints: an offline model…
A tutorial published by VideoDB instructs developers to keep video files out of app repositories and host them on VideoDB's service instead, uploading via the videodb npm package and embedding the resulting HLS stream or…
SageOx's Ryan Snodgrass released a new version of the company's session protocol that streams every coding-agent turn to teammates in under a second, replacing the prior model where a session reached the team only after …
Twilio published guidance recommending that teams evaluate customer-facing AI agents on production conversations rather than only pre-launch tests, citing a Twilio report finding that 86% of companies believe customers a…
Redis released a remote Model Context Protocol (MCP) server for Redis Cloud, hosted at https://cloud.mcp.redis.io/mcp, letting AI agents manage Redis Cloud subscriptions, databases and Agent Memory stores from within a c…
An open-source project called SHARD (shard-engine) released a distributed local LLM inference engine that pools GPU VRAM and system RAM across LAN machines to run 14B, 32B, and 70B+ open-weight models that no single card…
OpenAI released GPT-6.2 Astra despite behavioral regressions documented in its own system card, according to a November 17, 2026 post by Adam Holter on X. Holter wrote that "GPT-6.2 Astra is HORRIBLY behaved And OpenAI k…
Nvidia will stop selling its H200 chips in China, acting CEO Colette Kress said Tuesday, closing a pathway brokered by former CEO Jensen Huang and President Donald Trump that shipped chips through the U.S. under a 25% ta…
Amazon SageMaker HyperPod Inference now supports NVIDIA Dynamo, letting customers run multi-node and disaggregated prefill/decode LLM serving through the HyperPod Inference add-on, AWS announced. The add-on installs Dyna…
Microsoft released Microsoft-Decision-1.1 to general availability, a decision-oriented AI model for routing, classification, prioritization, verification and workflow control tasks, available through Microsoft Foundry, O…
Microsoft unveiled MAI Cyber 2 Flash at Ignite, a specialist security model from Microsoft AI that scores 92.3% on the CyberGym-E2E vulnerability-discovery benchmark versus 91.5% for GPT-6 Astra while using half the reas…
Microsoft unveiled a lineup of specialized AI models at its Ignite conference Tuesday, claiming MAI Code 2 delivers 96% of Claude Opus 5.5's performance on common enterprise coding tasks for 5% of the cost and 98% of GPT…
AirColibri 1.0.0, an Apache-2.0 Python stack for running the Qwen3.5 hybrid architecture (Gated DeltaNet plus attention) on consumer NVIDIA GPUs with insufficient VRAM, has been released on GitHub, with testing done on a…
OpenAI launched ChatGPT for Sales on November 17th, 2026, packaging a new Sales plugin with more than 30 new and enhanced partner plugins for account research, meeting preparation, proposals, follow-ups and pipeline revi…
A hack of 178 major websites including CNN, Zoom, IKEA and Equifax was traced to a security hole in Optimizely, a web tool large sites run behind the scenes to interface with AI, according to the New York Post. Separate …
Developer DIIIUA released ForgeMind 0.2.1-alpha, a GPL-3.0 native Kotlin Android AI workstation app, as a GitHub prerelease APK for Android 10+ (minSdk 29) with no store listing yet. ForgeMind supports five native LLM wi…
A developer ran a structured inference-tuning methodology on a dual-RTX-3090 Arch box serving a live coding lane on November 17, 2026, and found no tuning win that survived its own error bar. An initial A/B run produced …
Google updated Google AI Studio to build full-stack Android apps with Gemini, adding automatic Cloud Firestore provisioning with Security Rules and Google Sign-In, secure Gemini API calls via Firebase AI Logic and Fireba…
Korean AI startup VIDRAFT published an audit of Hugging Face's officially tagged benchmarks claiming its Darwin-180B model, built on a Recursive Self-Improvement (RSI) mechanism, took 12 first-place finishes across 57 le…
Google's Gemini 4 Carbon took the top spot on the MathArena math benchmark at 85.4% expected performance, but at an expected cost of $15.70 per run — roughly 15 times the $1.07 per run of OpenAI's third-place GPT-6.1 Sol…