AI Daily Briefing

Content is auto-organized by AI from public feeds for reference only. Verify important facts at the original source.

Today's attention is split between new model releases and infrastructure. Meta introduced two Muse Wonder models, while Microsoft unveiled its in-house Maia 300 with data center deployment planned from 2027. DeepSeek is reportedly set to lead a funding round for a robotics lab founded by a former Tencent AI scientist, signaling more capital moving toward embodied AI. On the regulatory side, NHTSA opened a defect review into Tesla Robotaxi speeding in school zones, and the UK plans to restore pre-release evaluations of leading models.

2026-10-13 AI briefing

1. NO.01NewSurge669↑ 780%

   Meta launches Muse Wonder and Muse Wonder Full models

   Meta launched two new models, Muse Wonder and Muse Wonder Full, claiming state-of-the-art results on personal and professional agent tasks and performance on par with the strongest available models. Muse Wonder is available now to Muse Code subscribers and through the Meta Model API, while Muse Wonder Full runs at maximum reasoning effort and stays subscriber-only. Meta says Wonder will later power its Muse Agent, without giving a timeline.

   WTNTBX+1Wired · AI, Techmeme, plus 5 more7 sources·7 reports
   • Wired: Meta's New Muse Wonder Model Puts It Back at the AI FrontierWired · AI · 10/13, 01:00
   • Meta Introduces Muse Wonder and Muse Wonder Full ModelsTechmeme · 10/13, 01:00
   • Meta Unveils New Series of Advanced AI Models to Rival OpenAI and AnthropicNYTimes · 科技 · 10/13, 01:00
   • Meta Launches Muse Wonder and Muse Wonder Full, Available to Muse Code SubscribersThe Verge · AI · 10/13, 01:00
   • Meta Says New Muse Model Reaches AI Frontier After 14-Month PushBloomberg Technology · 10/13, 00:56
   • Meta rolls out Muse Wonder and Muse Wonder Full on its Meta Model API platformX：TestingCatalog (@testingcatalog) · 10/13, 00:42
2. NO.02New555New

   Microsoft Unveils Maia 300, Deploying in Data Centers From 2027

   Microsoft has unveiled Maia 300, its third-generation in-house AI chip, built on TSMC's N2P process with a multi-die design, 288GB of HBM4 memory, 18TB/s bandwidth, 13.8 PFLOPS of FP4 compute and 950W SoC power. Deployment into production is planned from 2027, with racks holding up to 96 accelerators and clusters scaling to 12,000 chips. Microsoft says Maia 300 handles 40% more tokens per megawatt than leading GPU systems and delivers up to 7x the rack throughput of Maia 200, while it will keep buying Nvidia and AMD GPUs and has no plans to rent out Maia capacity.

   CTIC华+1cnBeta, The Decoder, plus 4 more6 sources·7 reports
   • Microsoft launches third-generation Maia AI chip, expands use of in-house siliconcnBeta · 10/13, 00:35
   • Microsoft plans to scale its own Maia AI chips to gigawatt levels by 2027The Decoder · 10/12, 23:24
   • Microsoft Unveils Maia 300 AI Chip With 288GB HBM4 and TSMC N2PIT之家 · 10/12, 22:45
   • Microsoft to Lean More on Homegrown AI Chips, Cutting Reliance on NvidiaCNBC · 科技 · 10/12, 22:04
   • Microsoft to deploy Maia 300 AI chips next year, targeting 40% inference gain华尔街见闻 · 10/12, 21:45
   • Microsoft unveils Maia 300, its third-generation in-house AI chip, for 2027 deploymentcnBeta · 10/12, 21:07
3. NO.03New533New

   DeepSeek reportedly to lead funding in robotics startup founded by ex-Tencent scientist

   DeepSeek is reportedly planning to lead a $400 million valuation funding round in a robotics large model startup founded by a former Tencent AI scientist, expanding its footprint into embodied AI.

   华T3CTB华尔街见闻, The Next Web, plus 4 more6 sources·6 reports

   DeepSeek
   • DeepSeek to lead Winverse AI seed round at a reported $400M valuation华尔街见闻 · 10/13, 00:19
   • DeepSeek to lead seed round in robot AI startup at about $400M valuationThe Next Web · 10/12, 21:22
   • DeepSeek to lead seed round in robotics AI startup Winverse AI at ~$400M valuation36氪 · 10/12, 20:58
   • DeepSeek Said to Lead Funding in Robotics AI Startup at About $400M ValuationcnBeta · 10/12, 20:40
   • DeepSeek Reportedly to Lead Seed Round in Robotic AI Startup Winverse AITechmeme · 10/12, 20:20
   • DeepSeek Backs Ex-Tencent AI Scientist’s $400 Million Robot LabBloomberg Technology · 10/12, 19:42
4. NO.04New191New

   NHTSA opens defect review into Tesla Robotaxi school-zone speeding

   NHTSA has opened a defect review in response to a petition alleging that some 2026 Tesla Model Y Robotaxi and Cybercab vehicles sped in school zones while autonomy was active. The petitioner provided three videos; the agency will assess whether to open a formal investigation. No crashes or injuries are documented, and about 1,000 vehicles may be affected.

   TBThe Next Web, Business Insider · Tech2 sources·2 reports
   • Tesla robotaxis face US safety review over school zone speedingThe Next Web · 10/13, 00:35
   • NHTSA Opens Defect Review Into Tesla Robotaxi School-Zone Speeding ClaimBusiness Insider · Tech · 10/12, 22:59
5. NO.05Surge181↑ 118%

   UK to restore pre-release evaluations for leading AI models

   According to Bloomberg Technology, the UK is restoring its authority to evaluate leading AI models prior to release, reversing previous policies that required White House priority approval. This adjustment changes the compliance path for cutting-edge AI model regulation in the UK.

   TBThe Next Web, Bloomberg Technology2 sources·2 reports
   • UK says it will again test OpenAI and Anthropic models before releaseThe Next Web · 10/13, 00:19
   • UK to Evaluate AI Models Before Release, Reversing US PolicyBloomberg Technology · 10/12, 19:05
6. NO.06New98New

   US nuclear power industry accelerates AI assistant adoption across 94 reactors

   The US nuclear power industry is accelerating the adoption of next-generation AI assistants into daily workflows to retrieve technical documents, process regulatory filings, analyze equipment fault records, and find solutions from decades of operating experience. The push covers 94 reactors and is driven by staffing shortages and maintenance challenges.

   CcnBeta1 source·1 report
   • US nuclear industry ramps up AI assistants across 94 reactorscnBeta · 10/13, 00:31
7. NO.07New98New

   Jensen Huang says Chinese models like DeepSeek, Kimi and GLM are good enough

   Speaking at a DBS forum, Nvidia CEO Jensen Huang said Chinese AI models such as DeepSeek, Kimi and GLM are already good enough. He described himself as a firm supporter of open source, calling it the best way to drive efficiency gains, and said the market will decide which models best fit different use cases.

   CcnBeta1 source·1 report

   DeepSeekKimi
   • Jensen Huang says Chinese AI models like Kimi, DeepSeek and GLM are 'good enough'cnBeta · 10/13, 00:37
8. NO.08New97New

   Microsoft proposes Fork-and-Flush to boost agent performance by up to 66%

   Microsoft researchers found that autonomous AI agents can get stuck in "idea basins" as growing chat context traps them in local solutions. Their proposed Fork-and-Flush method clears chat history while preserving the workspace, yielding up to 66% improvement under the same compute budget.

   TThe Decoder1 source·1 report
   • Microsoft's Fork-and-Flush: clearing chat history boosts agent results by up to 66%The Decoder · 10/13, 00:10