Daily AI News — August 14, 2026

🧠 AI Brain / Memory

• Agent memory goes mainstream as the "second brain" pattern hardens (mid-Aug 2026) — A wave of 2026 guides and benchmark reports (Mem0's State of AI Agent Memory 2026, Letta, Zep) converge on external, vector-backed memory as the fix for agents that forget context between sessions. Sources: Mem0, Value Add VC, sedim3nt (Substack)
• Cognee raises $7.5M seed to build memory infrastructure for AI agents (recent) — The startup is building a dedicated memory layer so agents can persist and retrieve structured knowledge across tasks, a sign investors are funding the "memory" primitive directly. Source: Cognee
• Digital Science launches Dimensions MCP servers (Aug 10, 2026) — Two Model Context Protocol servers give enterprise agents governed access to licensed research data, effectively wiring a curated knowledge base into agent memory. Source: Scienmag
• "Company second brain" framing spreads to teams (mid-Aug 2026) — New playbooks push shared, org-wide knowledge bases that both humans and agents can search, moving the second-brain idea from personal notes to enterprise memory. Sources: MindStudio, Taskade, Orbitant

🤖 AI Agents

• Nvidia unveils Nemotron 3.5 Lightning & NeMo Switchyard (Aug 13, 2026) — A 30B mixture-of-experts model claims up to 4× faster token generation with dynamic routing across open and proprietary models for agent workloads. Source: AI Agent Store
• River AI raises $1.1B (Aug 13, 2026) — xAI co-founder Igor Babuschkin's startup offers a training API doing LoRA fine-tuning and RL runs on open-weight models in 15–20 minutes. Source: AI Agent Store
• L&T launches AgenticIQ for enterprise (Aug 12, 2026) — The platform converts manufacturing and operations workflows into governed AI agents under defined governance boundaries. Source: AI Agent Store
• Meta releases Muse Glimmer (Aug 10–12, 2026) — An open 30B-parameter model that runs agent tasks on a single ~20GB GPU with Ollama and LM Studio support. Source: AI Agent Store
• Microsoft Intelligent Terminal 0.2 (Aug 12, 2026) — Adds per-tab agent selection, OpenCode integration, and local model support for developers building agentic workflows. Source: TechGenyz

🛠️ AI Tools

• Google releases Gemini 3.7 Flash (Aug 13, 2026) — A new fast flagship model aimed at low-latency, high-throughput use. Sources: LLM Gateway, AI Release Tracker
• Z.ai ships GLM-5.3 (Aug 14, 2026) — Latest frontier model addition to the release trackers. Source: AI Release Tracker
• DeepSeek-V4-Pro (0813) (Aug 13, 2026) — A professional variant of DeepSeek's V4 series, launched alongside reported API pricing increases. Sources: AI Release Tracker, AI Agent Store
• OpenAI GPT-5.6-Cyber & Grok 4.6 (Aug 10–12, 2026) — A security-specialized GPT variant and an updated Grok model both landed this week. Source: AI Release Tracker

🎓 AI Skills

• Demand for top AI skills more than doubles (Upwork report, 2026) — The most in-demand skills grew ~109% YoY, with AI Integration up 178% and AI Chatbot Development up 71% as AI embeds into everyday work. Source: Upwork / Barchart
• Creative AI skills surge fastest — AI Video Generation & Editing jumped 329% and AI Image Generation & Editing 95%, the steepest gains of any category. Source: Upwork
• Human–AI collaboration is the differentiator — The report finds pairing traditional expertise with AI proficiency boosts project completion by up to 70%, and 77% of leaders want specialized/fractional AI talent. Source: Upwork / Gloat
• Stanford AI Index 2026 & workforce trackers flag a widening AI skills gap, framing agent-building and AI integration as the highest-leverage capabilities to hire and train for. Sources: Stanford AI Index (Lightcast), Gloat, Coursiv

---

Sources: AI Agent Store, Agentic.ai, LLM Gateway, AI Release Tracker, Mem0, Cognee, Value Add VC, sedim3nt, Scienmag, TechGenyz, Upwork/Barchart, Gloat, Stanford AI Index (Lightcast), Coursiv, MindStudio, Taskade.