Meet Launa-Omni, Salesforce’s Voice Model for Real-Time Conversation
Salesforce AI Research announces Launa-Omni, a new voice model designed for real-time, two-way conversational interactions.
Salesforce AI Blog·19 Oct 2026
Use this view to inspect the underlying evidence corpus. For ranked developments, decision posture and interpretation, use Signals.
Raw feed or Signals?
Raw feed is chronological evidence. Signals ranks and interprets material change.
Salesforce AI Research announces Launa-Omni, a new voice model designed for real-time, two-way conversational interactions.
MIT Technology Review examines how web optimization for AI agents risks altering access dynamics for human internet users.
Hamilton Lane's Erik Hirsch states AI value creation remains heavily concentrated in private markets via venture, secondaries, and co-investments.
Anthropic plans to dedicate 5% of its shares to a public benefit fund to address concerns regarding AI-driven labor disruption.
Nscale faces investor pushback on its 30B dollar IPO valuation target, while Anthropic plans a public benefit fund for five percent of shares.
Cloudflare introduces Clairic, a System One extraction model designed to extract, compute, and forecast data from unstructured inputs.
Goodfire researcher Nick Cammarata published research claiming new insights into the underlying nature and structure of neural networks.
US financial institutions including Charles Schwab are actively blocking unauthorized third-party AI agents from accessing accounts.
Nvidia highlights cuTile, a Python-based DSL for GPUs aimed at accelerating state space models like Mamba-2.
TwelveLabs Pegasus 1.5 video language model is now available in Amazon Bedrock for structured video metadata generation.
Nvidia-backed startup doubleAI introduces code execution models designed to sit between software and silicon.
Deepgram ported its Nova-3 speech model to AWS Trainium2 using AWS Neuron and a Rust runtime to optimize batch transcription costs.
AWS published technical benchmarks showing a push-based NIXL connector for vLLM inference that improves latency metrics on EC2 instances.
AWS introduced bi-directional KV cache transfer in vLLM to lower time-to-first-token latency for multi-turn AI agents.
Meta detailed GenRecs Tokenization, a generative recommendation approach using stable model vocabulary for ad matching.
Google Cloud introduces MCP Apps for Gemini Enterprise, rendering live interactive UIs inside chat flows via sandboxed iframes.
Google Cloud highlights risks in LLM-as-a-judge evaluation frameworks, noting severe verbosity and position biases in automated scoring.
Proofpoint agreed to acquire Varonis for $6.3B, aiming to unify email, collaboration, and data security infrastructure.
Nvidia launched open AI models designed to support enterprise agent deployments in on-premise data centres and on PCs.
Meta published architecture insights on improving AI agent reliability by using thin orchestrators and specialized domain agents.
US congressional expertise gaps in artificial intelligence hinder the pace and technical depth of upcoming federal regulatory efforts.
The UK FCA announced plans to develop a regulatory framework covering agent identity, delegated authority, accountability and execution.
NorForge research reveals malicious actors using exposed AI-driven penetration-testing control panels for automated cyber operations.
Financial Times reports AI data centre designs may require blast-resistant concrete, jamming gear, and missile-defence protection.
Utah formed a digital ID consortium to pilot individual-controlled personal data sharing based on its state privacy law.
Fifth Third CEO Tim Spence downplayed competitive risks from AI agents, emphasizing traditional incumbent leverage during an earnings call.
Huntington Bank launched 1866 Studio, an AI-powered operating platform aimed at small-business customers.
TikTok and Shoplazza have partnered to integrate agentic commerce tools designed to automate product discovery and purchasing flows.
A hacker used autonomous AI agents to breach over 100 corporate websites, according to AI security firm Norforge and Reuters reports.
Anthropic plans to put 5% of its shares into a public good fund to offset AI job impacts as it prepares for an IPO.