OpenAI is rolling out Dots, always-on AI agents that run on their own cloud computers and learn from user feedback. The agents are accessible inside ChatGPT, Slack, and Teams, but only Pro and Business Premium subscribers got access starting September 29. The launch came roughly three weeks after Meta released its Muse personal agent, and the live demo ran into voice command problems. Team agents and other features are still on the roadmap.
Google has announced the Gemini 4 Argon model, also called Gemini 4 Pro, which can process 1.5 times more requests per second than Gemini 3.5 Pro while using less energy. The model remains experimental and is currently limited to select researchers and developers. Google says it is working to broaden access and improve the user experience.
OpenAI also shared details on the Jalapeño ASIC, a custom chip designed with extensive use of AI tools across hardware and software co-design. The chip moved from register-transfer level to tapeout in nine months, and the full concept-to-reveal timeline came in under two years. Richard Ho, OpenAI head of hardware, said this set a new industry baseline.
On the policy front, Spanish Prime Minister Pedro Sanchez spoke at Harvard's Institute of Politics about the need for AI regulation and third-party oversight. He pointed out that roughly 75 to 80 percent of computing capacity sits in the United States, while the EU holds only a single-digit share. A separate article argues that comparing frontier AI development to the Manhattan Project is misleading, since AI model weights can be copied at near-zero cost. That piece calls for continuous technical monitoring, air-gapped testing environments, and independent domestic supervisory authorities.
Nvidia CEO Jensen Huang told an audience in Washington on September 29, 2026, that the AI data center buildout in the US could create about one million jobs. He described it as the first major re-industrialization of the country in about 50 years, covering construction, power plant building, cooling systems, and pipe fitting. Huang noted the US is adding between 10 and 20 gigawatts of infrastructure each year. Meanwhile, Exiger CEO Jake Tapper urged more transparency and accountability after tech leaders including Google's Sundar Pichai and Microsoft's Satya Nadella pledged to self-police their AI products.
On the ground, three ABA therapy providers — 360 Behavioral Health, Acorn Health, and Verbal Beginnings — said AI tools help staff train faster and respond more effectively, with positive feedback from families. A separate report on AI in recruitment raised concerns about bias, transparency, and whether human oversight in hiring is meaningful or just performative. A reader also wrote about the growing presence of tools like Google Gemini and Amazon Q, warning that AI-generated content can erase artists' work and harm future careers.
Key Takeaways
- OpenAI launched Dots, always-on AI agents in ChatGPT, Slack, and Teams, initially for Pro and Business Premium subscribers starting September 29.
- Meta released its Muse personal agent about three weeks before OpenAI's Dots launch.
- Google announced Gemini 4 Argon (Gemini 4 Pro), which processes 1.5 times more requests per second than Gemini 3.5 Pro while using less energy, but it remains experimental.
- OpenAI's Jalapeño ASIC went from register-transfer level to tapeout in nine months, with a concept-to-reveal timeline under two years.
- Spanish PM Sanchez said 75 to 80 percent of computing capacity is in the US, while the EU holds only a single-digit share.
- An article argues AI governance should not mirror nuclear analogies and calls for continuous technical monitoring and independent supervisory authorities.
- Nvidia CEO Jensen Huang said the AI data center buildout could create about one million jobs, with the US adding 10 to 20 gigawatts of infrastructure annually.
- Exiger CEO Jake Tapper called for more transparency after Google's Sundar Pichai and Microsoft's Satya Nadella pledged to self-police AI products.
- Three ABA therapy providers reported that AI tools help staff train faster and have been well received by families.
- A reader warned that tools like Google Gemini and Amazon Q, along with AI-generated social media content, threaten artists' work and future careers.
OpenAI launches always-on AI agents called Dots in ChatGPT
OpenAI launched Dots, always-on AI agents that work inside ChatGPT. Each agent runs on its own cloud computer and learns from user feedback. Users can reach agents in ChatGPT, Slack, and Teams. Dots began rolling out to Pro and Business Premium subscribers on September 29, with some users waiting several days for access.
OpenAI's Dots agent launch faces early challenges
OpenAI launched Dots, AI agents that work around the clock across more than 4,000 apps. The launch came three weeks after Meta released its Muse personal agent. Most ChatGPT users cannot try Dots yet, since it is limited to Pro, Business Premium, and some Enterprise customers. The live demo had voice command problems, and more features like team agents are still coming.
OpenAI details AI-assisted design of Jalapeño ASIC
OpenAI's Jalapeño ASIC was designed with heavy use of AI tools, from hardware design to co-design with the software stack. The chip went from register-transfer level to tapeout in just nine months. The full timeline from concept to reveal was under two years. Richard Ho, OpenAI head of hardware, says this timeline set a new industry baseline.
Nuclear analogy poorly fits AI governance, article argues
An article argues that comparing frontier AI development to the Manhattan Project is misleading. Unlike nuclear weapons, AI model weights can be copied at near-zero cost and spread across commercial networks. The article calls for new oversight tools, including continuous technical monitoring, air-gapped testing environments, and independent domestic supervisory authorities.
Spain PM Sanchez speaks on AI and corruption at Harvard forum
Spanish Prime Minister Pedro Sanchez spoke at Harvard's Institute of Politics about AI regulation and corruption allegations. He said AI development needs rules, third-party oversight, and stronger reporting requirements. Sanchez noted that about 75 to 80 percent of computing capacity sits in the United States, while the EU holds only a single-digit share. The forum also turned to corruption claims involving members of his government.
AI is changing recruitment and raising privacy concerns
AI is reshaping how employers screen resumes, match candidates and conduct interviews. The session explores how personal data and automated recommendations can influence hiring outcomes. It also examines whether human oversight is truly meaningful or if recruiters simply defer to AI. Attendees will learn how to evaluate recruitment tools for bias, transparency and accountability.
ABA therapy providers see AI as a lasting tool
Three health care providers, 360 Behavioral Health, Acorn Health and Verbal Beginnings, discussed AI in behavioral health. They shared that AI can help staff train faster and work more supported. Families have also responded positively to AI tools. The providers offered tips for ethical AI use and change management in care settings.
Google announces Gemini 4 Argon AI model not yet available
Google announced the Gemini 4 Argon AI model, also called Gemini 4 Pro. It is designed to process 1.5 times more requests per second than Gemini 3.5 Pro and uses less energy. The model is still experimental and only available to select researchers and developers. Google says it is working to make it more accessible and user-friendly.
Exiger CEO cautious about tech execs self policing AI
Exiger CEO Jake Tapper reacted to tech leaders pledging to self-police their AI products. He said the pledge is a good start but called for more transparency and accountability in the AI industry. The pledge came from executives including Google's Sundar Pichai and Microsoft's Satya Nadella. Concerns raised include AI bias, job displacement and misuse.
Readers call for awareness of AI negative impacts
A 20-year-old reader wrote about growing up with the internet and watching AI spread into daily life. They pointed to tools like Google Gemini and Amazon Q and the rise of AI-generated social media content. The writer stressed that AI can erase the work of artists and harm future careers. They urged more people to recognize the dangerous impacts of AI.
Nvidia CEO says AI data center buildout could create 1 million jobs
Nvidia CEO Jensen Huang said the AI data center buildout in the US could create about one million jobs. He described it as the first major re-industrialization of the country in roughly 50 years. The work includes construction, power plant building, cooling systems, and pipe fitting. Huang noted the US is adding between 10 and 20 gigawatts of infrastructure each year. He spoke at an event alongside Elon Musk on September 29, 2026, in Washington.
Sources
- OpenAI launches Dots, always-on AI agents in ChatGPT with their own cloud computers
- OpenAI’s Muse Rival Arrives Half-Baked
- ‘This is how AI should be used’ — OpenAI head of hardware breaks down the AI-assisted design of its Jalapeño ASIC
- The Limits of the Nuclear Analogy in Frontier Artificial Intelligence Governance
- Spain Prime Minister Speaks on Future of AI, Corruption at IOP Forum
- Privacy, AI, and the new rules of recruitment
- “Here to Stay”: 3 Providers on The Future of AI in ABA Therapy
- Google announces Gemini 4 Argon AI model, but you can't use it yet
- CEO reacts to tech execs pledging to ‘self-police’ their AI products
- Readers speak: We all need to know the negative impacts of AI
- Nvidia CEO Jensen Huang: AI data center buildout could create 1 million jobs and reshape the US economy
Comments
Please log in to post a comment.