OpenAI's Codex is helping companies like Cult.fit cut costs and build internal tools without dedicated engineering teams. The company has created over 20 applications using the OpenAI API, saving more than $100,000 a year. Customer support costs have dropped by more than half thanks to AI-driven tools, showing how accessible AI is becoming for non-technical workflows.
Meanwhile, the global race around AI continues to intensify. Chinese companies are rolling out increasingly powerful models, competing directly with the United States. Experts from Janes and the Council on Foreign Relations say the question is no longer just about capability, but about who gets to shape the future rules of AI development.
That rivalry was on display at the September 2026 summit between President Trump and President Xi Jinping in Washington. The two leaders set a November 2026 deadline for a technical dialogue on AI and agreed to establish a bilateral AI incident hotline. Still, differences persist, including the US term super intelligence versus China's artificial intelligence, and progress will depend on bringing industry experts into the conversation.
On the privacy front, Anthropic's Claude Code has drawn scrutiny after it was found to send a user's email address in the User-Agent header when fetching websites. Anthropic says the cc-plugin-privacy tool blocks that substitution, but critics argue the email remains in context by default. Users are advised to verify the plugin is enabled and pin allowed User-Agent emails to approved domains until the issue is resolved.
In other AI news, Mauritius Telecom partnered with Amazon Web Services to launch the myt Learning Hub Portal, offering free digital and AI training through 16 courses. The platform supports the government's goal to train 50,000 Mauritians, with officials calling it a national strategic imperative for AI literacy. Cisco also revealed it has built a full-stack enterprise AI factory across its global data centers to tackle low GPU utilization and rising cloud costs, serving a workforce of 90,000 through its super agent MyAgent.
Looking ahead, researchers have developed an AI model called LUMINA to help plan future space outposts on the Moon, Mars, and asteroids. Built on a lunar vision-language foundation model, LUMINA will be tested with new lunar data supported by NASA's Artemis program. Separately, a review in Discover Networks finds that 6G networks will need AI optimization and green radio hardware to reach carbon neutrality, though the proposed frameworks still need real-world testing.
Key Takeaways
- <ul><li>Cult.fit built over 20 internal apps using OpenAI's Codex, saving more than $100,000 annually and cutting customer support costs by over 50 percent.</li><li>The China-US AI rivalry continues to escalate, with experts debating who will shape the future rules of AI development.</li><li>The September 2026 summit between President Trump and President Xi Jinping set a November 2026 deadline for a technical AI dialogue and a bilateral AI incident hotline.</li><li>Anthropic's Claude Code exposed user email through User-Agent headers, with privacy reliant on the cc-plugin-privacy tool rather than a built-in fix.</li><li>Mauritius Telecom and Amazon Web Services launched the myt Learning Hub Portal with 16 free AI courses, aiming to train 50,000 Mauritians.</li><li>Cisco deployed a full-stack AI factory across its global data centers to address low GPU utilization and rising cloud costs for its 90,000-person workforce.</li><li>DOGE Acting Administrator Amy Gleason envisions AI chatbots replacing individual agency websites, citing a 2.6 percent drop in government service satisfaction.</li><li>Kevin Roose's book The AGI Chronicles explores the accelerating race to build artificial superintelligence and the concerns within the AI community.</li><li>LUMINA, a lunar vision-language AI model, could help plan future space outposts and will be tested with NASA Artemis program data.</li><li>A Discover Networks review finds 6G networks require AI optimization and green radio hardware for carbon neutrality, but proposed frameworks still need real-world testing.</li></ul>
Cult.fit uses Codex to build tools and cut costs
Cult.fit is using Codex and the OpenAI API to help non-engineering teams build internal applications. The company has built over 20 applications and saved more than $100,000 annually. Customer support costs have dropped by more than 50 percent through AI-driven tools.
AI and green radio hardware key to carbon-neutral 6G
A review published in Discover Networks finds that 6G networks will need AI optimization and green radio hardware to reach carbon neutrality. The study introduces frameworks like GRTEM, HYASPA, and Eco-DTA to balance energy use and environmental impact. The authors say these models still need real-world testing through simulations and 6G test environments.
AI system LUMINA could help plan future space outposts
Researchers developed an AI model called LUMINA to help mission planners map out future space outposts. The system is built on a lunar vision-language foundation model and is designed to support decisions for missions on the Moon, Mars, and asteroids. The team plans to test LUMINA on new lunar data, supported by NASA's Artemis program.
China-US AI arms race continues to escalate
Chinese companies are rolling out increasingly powerful AI models, intensifying competition with the United States. The discussion focuses on who will shape the future rules of AI development. Experts from Janes and the Council on Foreign Relations joined the conversation to analyze the escalating rivalry.
Mauritius Telecom launches free AI skills platform
Mauritius Telecom launched the myt Learning Hub Portal with Amazon Web Services to offer free digital and AI training. The platform includes 16 courses and supports the government's goal to train 50,000 Mauritians. Officials called the initiative a national strategic imperative for AI literacy across all communities.
Kevin Roose on the accelerating race to build AI superintelligence
Kevin Roose explains why AI insiders are worried about the rapid pace of development. In his book The AGI Chronicles, he discusses the race to create artificial superintelligence. He spoke with Scientific American about his research on the AI boom and the extraordinary advances of the last decade.
Cisco builds enterprise AI factory to scale performance and security
Cisco IT deployed a full-stack AI factory across its global data centers to overcome scaling challenges. The company faced low GPU utilization at 10-15 percent, rising cloud costs, security risks, and power density limits. Its enterprise super agent MyAgent serves a workforce of 90,000 people. The new infrastructure pools compute, networking, and security to deliver high performance at enterprise scale.
DOGE leader envisions AI-driven future for federal services
DOGE Acting Administrator Amy Gleason says AI chatbots could one day replace individual agency websites. She spoke at an event about modernizing America.gov and improving digital government services. Satisfaction with government services fell 2.6 percent over the past year according to the American Customer Satisfaction Index. Gleason warned that agencies lack talent and tools to keep up with the pace of technological change.
China and USA build on AI progress after Trump-Xi summit
The September 2026 summit between President Trump and President Xi Jinping in Washington produced a November 2026 deadline for a technical dialogue on AI. The two leaders also agreed to establish a bilateral AI incident hotline. Differences remain, including the US term super intelligence versus China's artificial intelligence. Progress will depend on involving industry experts and addressing mutual dependencies in semiconductors and critical minerals.
Claude Code exposed user email through User-Agent headers
Claude Code sent a user's email address in the User-Agent header when fetching websites. Anthropic says the cc-plugin-privacy tool blocks that substitution but critics argue the email remains in context by default. Users are advised to verify the plugin is enabled and pin allowed User-Agent emails to approved domains. Until the email is removed from context by default, privacy depends on a plugin rather than a built-in guarantee.
Sources
- Cult.fit builds operational muscle with Codex
- AI and green radio hardware may hold the key to carbon-neutral 6G networks, review finds
- Could next-gen artificial intelligence help map out future space outposts?
- How the China-US AI arms race is escalating
- Mauritius Telecom launches free digital and AI skills training platform
- Kevin Roose says the race to build superintelligence is accelerating
- How Cisco Built an Enterprise AI Factory: Scaling Economics, Performance, and Security
- DOGE leader outlines AI-driven future for federal service delivery
- Building On Progress In China–USA Talks On AI
- Claude Code sent your email in User-Agent
Comments
Please log in to post a comment.