Models & Labs
ChatGPT mobile gets voice-based agentic Work tab
TechCrunch AIhigh confidence
Why it matters
- →Brings agentic workflow execution (drafting, summarizing) to mobile via voice for paying users.
- →Enables seamless cross-device session resumption between mobile and desktop.
- →Signals OpenAI's push to make voice the primary interface for complex AI tasks, not just simple queries.
OpenAI announced Wednesday that voice-based agentic features are now available in the ChatGPT mobile app for Plus and Pro subscribers. Users can access the 'Work' tab to perform tasks such as drafting emails, summarizing Slack messages, building websites, or accessing financial data using voice commands. The update includes richer text output from voice conversations and allows users to switch between text and voice modes while resuming sessions across devices. This rollout follows the July introduction of GPT-Live on desktop and aims to capitalize on growing user demand for voice-driven AI assistants capable of handling complex, multi-step tasks.
Read originalMore from TechCrunch AI
© TechCrunch AIInvestment · $311M
Market & Regulationbusiness
Enveda raises $311M for AI drug discovery
Enveda’s $311 million Series E signals that the biotech market is betting heavily on nature-derived AI drugs. By doubling its valuation to $2 billion, investors are backing a strategy that hunts for medicines in plants and microbes rather than synthesizing them from scratch. With candidates already in clinical trials for skin conditions and GLP-1 weight loss maintenance, Enveda moves beyond theoretical promise into human testing. This funding validates the niche of using AI to decode complex natural compounds, a space where few players have reached this stage.
TechCrunch AI
© TechCrunch AIVideo & Creative AIother
YouTube Music adds conversational AI and podcast guides
YouTube Music is finally catching up to Spotify’s AI ambitions with two features that move beyond simple search. Ask Music lets users build queues or dig into music history using natural language, tapping into a catalog of over 300 million tracks. Meanwhile, Your Podcast Lineup generates spoken previews to cut through the noise of endless show recommendations. This isn't a breakthrough in model capability, but it signals that major platforms are standardizing conversational interfaces as the primary way to consume media.
TechCrunch AI
© TechCrunch AIGeneral AIother
YouTube launches AI-powered custom feeds
YouTube is finally letting users curate their own discovery streams using natural language prompts powered by Gemini. This moves beyond simple keyword filtering into semantic understanding, allowing for nuanced requests like 'relaxing commentary' or specific commute contexts. It aligns YouTube with Bluesky and Threads in the race to democratize feed curation, acknowledging that one-size-fits-all algorithms no longer satisfy diverse user intents. The feature pins these custom streams to the home tab without disrupting the main recommendation engine, offering a parallel layer of control over the platform's massive 20 billion-video library.
TechCrunch AI
More in Models & Labs
Models & Labscoding
vLLM v0.30.1rc0 adds AMD MI355 NVFP4 support
vLLM is quietly closing the hardware gap for AMD users with this release candidate. By adding dense NVFP4 and MoRI kernel mirrors for the new MI355 GPU, they are enabling high-efficiency inference on hardware that previously lacked first-class support. This isn't just a driver update; it's a critical infrastructure patch that allows enterprises to deploy advanced quantization formats on AMD silicon without waiting for upstream integration. The inclusion of OpenAI Codex in the commit history suggests automated testing is helping maintain this parity, making AMD a more viable option for cost-sensitive inference workloads.
vLLM Releases
Models & Labsother
llama.cpp b11149 adds backend testing and ROCm 10 support
This release quietly solidifies llama.cpp’s position as the universal inference runtime by adding explicit ROCm 10.0 builds for both Linux and Windows. The inclusion of CUDA 13.4 alongside the existing 12.x variants ensures compatibility with the latest NVIDIA driver stacks without forcing users to stick to older libraries. More importantly, the new backend testing infrastructure means these diverse hardware configurations are now validated systematically rather than left to chance. This reduces fragmentation for developers running on AMD or newer NVIDIA cards who previously had to troubleshoot build issues manually.
llama.cpp Releases
Models & Labsother
llama.cpp b11154 improves testing and adds CUDA 13
This release refines the internal test suite for better visibility, but the real signal is platform expansion. CUDA 13 builds are now available across Linux and Windows, giving developers early access to the latest NVIDIA stack without waiting for stable drivers. AMD ROCm support also advances with version 10.0 binaries, keeping pace with hardware shifts. While KleidiAI on Apple Silicon is currently disabled, the core inference runtime remains robust across major architectures. This is a maintenance-heavy update that ensures compatibility with cutting-edge GPU libraries.
llama.cpp Releases