SourceUpdated 2w ago · 462 words

Daily AI News — August 15, 2026

Covering the last ~48 hours (Aug 12–14, 2026).

🧠 AI Brain / Memory

  • Starling publishes "Universal Cognitive Architecture" open standard (Aug 13) — An open standard for connecting organizational knowledge to AI models so companies can use stored knowledge without losing control of it. (Research Live)
  • DeepJudge launches Agent Handoff Protocol (Aug 13) — An open standard that lets users move between AI products while preserving full context, with Harvey and Thomson Reuters backing adoption. (LawSites)
  • Tencent Cloud adds "Team Memory" to its open-source Agent Memory project (Aug 12) — Lets agents share conversations, documents, code, and institutional knowledge across a team. (Taiwan News)
  • Mem0 releases "State of AI Agent Memory 2026" benchmark (this week) — New evaluations comparing persistent-memory stacks (Mem0, Letta, Zep) as agent memory becomes its own infrastructure layer. (Mem0)

🤖 AI Agents

  • Writer releases Palmyra X6 (Aug 14) — Flagship model runs multistep agent workflows at 52% lower cost and 48% faster than its predecessor. (aiagentstore.ai)
  • FriskAI launches runtime intelligence for agents (Aug 14) — A $3.6M-backed startup providing audit trails and behavioral analytics for production AI agents. (aiagentstore.ai)
  • SpaceXAI opens Grok Bot beta (Aug 13) — Persistent agents that sign into apps and websites and keep executing multistep jobs after the user disconnects. (aiagentstore.ai)
  • Okta unveils a framework to secure enterprise AI agents (this week) — New identity/authorization framework aimed at governing autonomous agents in production. (TechRadar Pro)

🛠️ AI Tools

  • Google launches Gemini 3.7 Flash (Aug 14) — Optimized for coding and agent tasks at roughly half the price of Gemini 3.6 Flash. (aiagentstore.ai)
  • OpenAI ships GPT-5.6 + "Sol" Ultrafast API tier (Aug 14) — Platform upgrade paired with a new tier delivering up to 14× faster throughput. (OpenAI News)
  • Nvidia releases Nemotron 3.5 Lightning (Aug 13) — A 30B-parameter model with up to 4× faster output generation than comparable systems. (aiagentstore.ai)
  • Meta releases Muse Glimmer (Aug 12–13) — A 30B open-weight model that runs capable agents locally on a single consumer GPU. (aiagentstore.ai)

🎓 AI Skills

  • Upwork: demand for top AI skills more than doubles (2026 report) — AI-enabled skill demand up 109% YoY, led by AI Video Generation & Editing (+329%), AI Integration (+178%), and AI Data Annotation (+154%). (Upwork / Barchart)
  • UK government launches AI bootcamp workforce initiative (Aug 14) — National training programs to move workers into AI-skilled technical roles, though analysts flag a security-awareness gap. (The Guardian)
  • Stanford AI Index 2026 skills data (this week) — Fresh labor-market analysis on where AI job postings and skill requirements are concentrating. (Lightcast / Stanford HAI)

Sources: aiagentstore.ai, agentic.ai, Research Live, LawSites, Taiwan News, Mem0, TechRadar Pro, OpenAI News, The Guardian, Upwork/Barchart, Stanford AI Index (Lightcast), techmaniacs.com.