https://audio-briefings.pages.dev/tech-mornings/ en-us A daily briefing on AI developments, new capabilities, what's around the corner, and the security news you need to know. A daily briefing on AI developments, new capabilities, what's around the corner, and the security news you need to know. no episodic DeepSeek V4.1 Flash (Sept): 552B-parameter MoE, only ~8B active per token, new architecture cuts per-token memory to ~1/4; pricing $0.30/$1.20 per M tokens, halved off-peak; since mid-Sept DeepSeek routes V4 Pro requests to Flash at Flash rates — undercutting its own flagship Inception Mercury 2.5 (Sept): diffusion-based language model — generates many tokens at once then refines; 1,107 tokens/sec on standard Nvidia GPUs; 260K context; 80% launch discount ($0.04/$0.15 per M tokens); targets voice agents/search/RAG latency New Muse abilities: none — no new Muse launch since Sept 29 Small Busine Sat, 03 Oct 2026 14:28:21 GMT https://audio-briefings.pages.dev/tech-mornings/episodes/tech-morning-briefing-saturday-october-3-2026-2026-10-03/ 228 Google DeepMind announced Gemini 4 Argon (Sept 30): frontier model rolled out first to cyber defenders via the Fairwind program (unguarded version); benchmarks: 77.9% DeepSWE v1.1, Text Arena top at 1525 (20 clear of Opus 4.6 High), 1M-token output (up from 64K); pricing $2/$10 per M tokens intro, 95% off cached inputs, rates double when intro ends; Bloomberg-reported caveat from employees that real coding work lags benchmarks, disputed by Google Cohere Embed 5 (Oct 1): new family of embeddings models powering enterprise search/retrieval New Muse abilities: no brand-new Muse launch since Sept Fri, 02 Oct 2026 14:29:43 GMT https://audio-briefings.pages.dev/tech-mornings/episodes/tech-morning-briefing-friday-october-2-2026-2026-10-02/ 226 Gemini 4 Pro: leaked checkpoint points to October launch; DeepMind shipped Lyria 3.5 full-length AI music model in Gemini app/API Dreamforce 2026 wrap: Salesforce AIforce (data/workflows/permissions/logic exposed to outside AI interfaces); Koa CRM reasoning model for multi-step Agentforce tasks; Claudeforce (Salesforce context in Claude); Nvidia NeMoTron-based models into Missionforce for select customers in October 2026 Open source: DeepSeek-V4-Flash-Vision-Exp (first native vision model, MIT license, matches text model on agents/reasoning, adds image input); Alibaba Qwen3.8-Max "0902" tops A Thu, 01 Oct 2026 14:51:23 GMT https://audio-briefings.pages.dev/tech-mornings/episodes/tech-morning-briefing-thursday-october-1-2026-2026-10-01/ 176 Good morning, and welcome to the very first Tech Morning Briefing for Wednesday, September 30th. I'm Maya. And I'm Devon. Huge day in A I. OpenAI just wrapped its DevDay in San Francisco with more than twenty launches, Anthropic shipped its second new model in under a week, and Meta's own Muse keeps gaining new abilities. Plus a rough week in security. Let's get into it. Start with OpenAI's DevDay… Wed, 30 Sep 2026 18:15:10 GMT https://audio-briefings.pages.dev/tech-mornings/episodes/tech-morning-briefing-wednesday-september-30-2026-2026-09-30/ 243