Nvidia Reportedly Building Trillion-Parameter Open-Source AI Model ‘Nemotron 4’ to Take on the Best in Class
August 11 — Reports are out that Nvidia is quietly working on a next-gen open-source AI model called Nemotron 4, and here’s the kicker: it’s expected to pack at least one trillion parameters. That puts it squarely in the ring with the world’s top open-source models. The play here is pretty clear — Nvidia wants to grow the open-model ecosystem to expand where AI gets used, which in turn fuels even more demand for its GPUs.
This isn’t Nvidia’s first rodeo with open-source. They’ve already shipped the Nemotron series, including the Nemotron-4 340B lineup with 340 billion parameters, which is mainly used for generating training data for large language models, supporting model development, and powering enterprise AI applications.
Google DeepMind Team Warns Applicants: Our AI Screening Might Overlook Your Resume — Here’s the Workaround
August 11 — A team from Google’s DeepMind, specifically the AGI Safety and Coordination group, is telling job seekers that their internal recruiting system has a real chance of botching resume screening. Their advice? When you submit an application, also fill out the extra form so a human on the team can actually see your resume. It’s a candid admission that the AI filter isn’t perfect — and a helpful heads-up for anyone trying to get noticed.
BAAI Research Institute Unveils AREX, an Autonomous Research Agent
August 11 — The Beijing Academy of Artificial Intelligence (BAAI) has launched AREX, an autonomous research agent that introduces a pioneering “research → verify → re-research” dual-loop recursive framework. This lets the AI keep correcting itself during long, complex tasks, progressively zeroing in on the right answer. The model weights are already open-sourced, and a research-oriented trial version is live, offering features like news tracking, paper filtering, and podcast summaries — think AI that reads papers on its own, follows trends, and whips up literature reviews for you.
Zuckerberg: Meta Will Make ‘Personal Superintelligence’ Mainstream and Push Back Against AI Power Hoarding
On August 10, local time, Meta CEO Mark Zuckerberg published a lengthy post titled “The Future Belongs to Everyone,” saying Meta is committed to making “personal superintelligence” widely accessible while continuing to champion open-source AI models. His argument? Open models prevent AI capabilities from being concentrated in too few hands, giving more people and businesses access to cutting-edge AI tools.
Zuckerberg stressed that the current open-source ecosystem is “strong,” and clamping down on open-source AI would only accelerate centralization — bad for both security and economic growth. He also revealed that Meta’s Superintelligence Lab is already up and running, and that the company will resume releasing some open-source models down the road.
He also weighed in on distillation in AI model training, arguing that models learning from other models is a core principle of the open-source ecosystem and shouldn’t be casually dismissed as harmful.
Nvidia Teams Up with Blackstone, Goldman Sachs and Four Other Wall Street Giants to Build a $500B AI Infrastructure Funding Platform
August 10 — Nvidia announced partnerships with Apollo Global Management, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to create financing platforms for AI computing infrastructure, aiming to mobilize over $500 billion in third-party capital. Nvidia CEO Jensen Huang said in a statement: “We’re bringing together the world’s leading long-term capital providers to independently evaluate and finance AI infrastructure projects. These platforms will help customers access massive amounts of scarce compute and build AI factories that will power industries and nations across the AI era.”
Li Auto’s Second-Gen AI Glasses Expected to Hit the Market in the Second Half of Next Year
August 10, evening — Li Auto, in a Q&A session about its AI glasses brand Livis, revealed that the second-generation Li Auto AI Glasses Livis is planned to arrive in the second half of next year. A native adaptation based on HarmonyOS is also in the works. Meanwhile, the new Li L9, L8, and L6 now support AI-glasses-guided parking. More models are being tested, and by mid-to-late August, the feature will open up to a broader set of users.
Anthropic Reportedly Strikes a $9B+ Compute Deal with Riot Platforms
According to reports, Anthropic has signed a $9.1 billion compute agreement with Riot Platforms, a bitcoin mining and digital infrastructure company. Riot disclosed on August 10 that it has entered a 20-year deal to provide 191 megawatts of computing capacity from its Rockdale campus in Texas to a “leading frontier AI company.”
DaiMeng Robotics Closes Hundreds of Millions in Funding, Led by Ant Group
August 11 — DaiMeng Robotics announced it has completed a funding round worth hundreds of millions of yuan, led by Ant Group with strong follow-on participation from existing shareholders. Within the last two months, the company has closed a Series A of 100 million yuan and a strategic round of several hundred million yuan. Previous investors include Inovance Industry Investment, China Mobile’s Chain Leader Fund, Lenovo Capital, China Merchants Venture Capital, and China Telecom. DaiMeng Robotics was incubated out of a Hong Kong University of Science and Technology research team, officially began operations in 2023, is headquartered in Shenzhen, and also runs an R&D center in Hong Kong. The company’s goal is to build a full-stack physical AI that connects tactile sensing, data, and models.
Xiaomi Kicks Off 2027 Campus Recruitment, AI Roles Up 50%
August 10 — Xiaomi Group announced the official launch of its 2027 global campus recruitment drive, targeting graduates who will complete their studies between September 1, 2026, and August 31, 2027. The recruitment spans more than 17 job categories, with AI-related positions seeing a 50% increase in demand.
