🏛️ ChatGPT, other AI chatbots approved for official use in US Senate, NYT reports ↗
The U.S. Senate has reportedly greenlit ChatGPT, Gemini, and Microsoft Copilot for official use by aides. That marks a significant symbolic threshold - AI tools are no longer perched at the fringes of government work, but being folded directly into it. (Reuters)
What stands out is how ordinary it already sounds. Not some moonshot pilot, not a glossy demo - just chatbots being wired into Senate platforms, as though this had long been the natural destination. (Reuters)
⚙️ AI startup Thinking Machines clinches capital and a major chip supply deal from Nvidia ↗
Mira Murati’s Thinking Machines Lab locked in a multiyear partnership with Nvidia, along with a significant investment and access to at least one gigawatt of next-gen processors. That is not a standard startup arrangement - it is industrial-scale, almost utility-sized. (Reuters)
The plan is to use Nvidia’s upcoming Vera Rubin systems to train its models. The story feels less like a single funding event and more like a contest over who gets compute now - because in AI, chips are starting to resemble territory. (Reuters)
⚡ France to harness nuclear power for AI data centres, says Macron ↗
France is pitching its nuclear fleet as an AI advantage, with Emmanuel Macron saying the country can open data centers and meet AI demand thanks to large exports of decarbonized electricity. It is a distinctly French flex - not just bigger models, but cleaner power behind them. (Reuters)
The subtext is clear enough: AI competition is turning into an energy competition too. Compute is hungry, data centers are hungrier, and now national power grids are being pulled straight into the middle of the story. (Reuters)
🪖 Google to Provide Pentagon With AI Agents for Unclassified Work ↗
Google is rolling out Gemini AI agents for Pentagon staff, initially on unclassified networks, to automate routine work across a massive workforce. That may sound administrative, even drab - but it is one of those low-key shifts that can end up carrying real weight. (Bloomberg)
These are agent-style systems, not just chat windows with improved manners. They are meant to act on tasks independently, nudging military bureaucracy one step closer to AI-assisted operations, even if the first use case is paperwork rather than anything dramatic. (Bloomberg)
📄 Google rolls out new Gemini capabilities to Docs, Sheets, Slides, and Drive ↗
Google is pushing Gemini deeper into Workspace, adding tools that can generate formatted drafts, slides, and spreadsheets using info pulled from Gmail, Chat, and Drive. It is the classic office-suite play, certainly, but also a quiet attempt to make AI feel less like a separate app and more like the default way work gets done. (TechCrunch)
The interesting part is the stitching together of your own work context. Not just "write me something," but "write me something from all the material I already forgot was scattered across my accounts" - which is practical, faintly eerie, and very on-brand for this era. (TechCrunch)
🛒 Judge orders Perplexity to stop AI agents from shopping on Amazon ↗
A federal judge ordered Perplexity to stop its AI agents from making Amazon purchases on users’ behalf, after finding Amazon had shown strong evidence of unauthorized access. So yes, the agent economy ran into an actual wall - the legal one. (The Verge)
The ruling also says Perplexity must destroy any Amazon data it may have obtained, with the injunction taking effect after a short window for appeal. It is a sharp reminder that "agentic" sounds futuristic until it collides with platform control, account access, and old-school gatekeeping. (The Verge)
🎭 Zoom introduces an AI-powered office suite, says AI avatars for meetings arrive this month ↗
Zoom unveiled an AI office suite with docs, slides, sheets, an agent builder, meeting translation, and photorealistic avatars for meetings. That is a lot at once - perhaps too much at once - but the company clearly wants to be more than the app you open because someone sent a link. (TechCrunch)
The avatars are the part people will remember. They are designed to stand in for you when you are not camera-ready, which is either convenient or a little dystopian, depending on your caffeine levels. Likely both. (TechCrunch)
FAQ
Why were ChatGPT, Gemini, and Copilot approved for use in the U.S. Senate?
The article suggests this matters because AI tools are no longer being treated as side experiments. They are being accepted as standard software for official staff work. The approval is as symbolic as it is practical, signaling that chatbot-style AI is moving into routine government workflows rather than remaining at the margins.
What does the Senate decision say about where AI is heading next?
It points to AI becoming part of standard institutional infrastructure. Instead of flashy pilots or isolated trials, the story presents these tools as something that can be integrated into existing platforms and daily work. That shift makes AI adoption appear less experimental and more like an expected layer of modern administrative systems.
Why is Thinking Machines’ Nvidia deal such a big AI story?
Because it appears far larger than a typical startup funding update. The company secured both capital and a multiyear chip supply arrangement tied to next-generation systems. In the article’s framing, that makes compute access feel like a strategic asset, where securing enough processors is becoming central to who can build leading AI models.
How are chips becoming a competitive advantage in AI?
The piece describes chips almost like territory: something scarce, strategic, and worth locking down early. Thinking Machines plans to use Nvidia’s Vera Rubin systems, which shows how model development now depends not just on talent or ideas but on long-term access to massive compute. In many AI pipelines, hardware supply can determine what is realistically possible.
Why is France using nuclear power as part of its AI strategy?
France is presenting its electricity system as an advantage for AI data centers. Macron’s point is that the country can support growing compute demand while drawing on large supplies of decarbonized power. The broader implication is that AI competition is no longer only about models and chips; it is also about who can reliably power large-scale infrastructure.
What is changing with AI agents in government and defense work?
The article highlights that Google is providing Gemini-based agents for Pentagon staff on unclassified networks. That matters because agents are presented as tools that can carry out tasks, not just answer prompts in a chat box. Even when the initial work is routine or administrative, it signals a move toward more active AI-assisted operations inside large institutions.
How is AI getting built into everyday work apps like Google Workspace and Zoom?
Both examples show AI being folded directly into office software people already use. Google is expanding Gemini across Docs, Sheets, Slides, and Drive with context drawn from existing work materials, while Zoom is adding docs, spreadsheets, agent tools, translation, and avatars. A common approach is to make AI feel less like a separate destination and more like the default workflow layer.
What does the Perplexity-Amazon court fight reveal about AI agents?
It shows that the agent model runs into hard limits when it intersects with platform rules, account access, and data rights. According to the article, a judge ordered Perplexity to stop Amazon shopping activity and destroy related data it may have obtained. That makes the legal side of AI agents seem just as important as the product experience.