👋 Good morning. This week’s AI developments cut straight to the core of where the industry is heading. Infrastructure is shifting, policy is being stress-tested, and major players are pushing AI deeper into everyday decisions—from how models access the web to how governments structure national research. The stories here aren’t hype pieces; they show the pressure points shaping the next phase of AI adoption, and the consequences when things break down.
🗨️Deloitte’s AI Shortcut
It’s not just students getting caught using AI to write assignments, now Deloitte is in trouble. A report by Deloitte, commissioned by a Canadian province — wasn’t just miscalculated: it was built, in part, on research that never existed. A 526-page “health-care workforce plan” that cost nearly CAD 1.6 million reportedly contains multiple false citations, referencing academic papers so fictitious that listed authors deny ever publishing them.
That’s more than AI writing a student essay. This is AI, or AI-assisted work, being trusted to shape public policy.
The document was meant to guide policy on virtual care, staffing shortages, and post-pandemic workforce incentives. Instead, some of its foundational “research” turns out to be fake. Some citations cite real researchers who say they had no part in the papers; others refer to studies that can’t be found in the referenced journals.
When a high-stakes, multi-million-dollar government report is built on AI-generated or fabricated research, it doesn’t just undermine one study, it puts entire policy decisions, public trust, and resource allocations on a shaky foundation.
🚀Parallel Web Systems is redefining web search for AI agents
An AI startup from the U.S., Parallel Web Systems, is building the web for AI agents. Founded by former Twitter CEO Parag Agrawal, the company raised $100 million in November 2025, giving it a valuation of $740 million.
Parallel is focused on a very different kind of search: instead of returning links for humans, its APIs deliver optimized content “tokens” that AI models can directly use. This approach helps reduce errors, improve accuracy, and cut operational costs for businesses that rely on AI to analyze data, write code, or assess risk.
The funding will be used to accelerate product development, acquire more customers, and tackle one of AI’s thorny challenges: web content increasingly locked behind paywalls. Parallel is even exploring new ways to encourage publishers to keep content accessible to AI systems.
Why it matters: As AI agents become primary users of the web, tools like Parallel hint at a future where AI doesn’t just read the web, it interacts with it efficiently and reliably. The startup shows that the next wave of AI innovation isn’t just smarter models, but infrastructure that lets those models actually get useful work done.
🔨AI Tools and updates: Shop With ChatGPT!
ChatGPT got a new feature: Shopping Research, a built‑in "AI shopper" designed to change how people hunt for products and gifts.
When you tell ChatGPT what you need, say, “quiet cordless stick vacuum for a small apartment” or “a gift under $100 for a friend who likes photography”, ChatGPT will ask clarifying questions, search the web across trusted retailers and review sites, and deliver a personalized buyer’s guide with top picks, trade‑offs, and up‑to‑date specs, reviews, and price/availability info.
The feature just launched for all logged‑in users (Free, Go, Plus, Pro) on web and mobile. To help during holiday season shopping, usage is “nearly unlimited.”
Behind the scenes, Shopping Research runs on a GPT‑5‑mini model fine‑tuned for product discovery: optimized to read multiple sources, cite them, and synthesize accurate, up‑to‑date product information — something standard chat models struggle with when faced with detail‑heavy decisions.
What’s interesting:
It marks a shift for ChatGPT, from general helpful assistant toward decision‑making assistant. Instead of just answering questions or summarizing info, ChatGPT tries to solve user problems: “what should I buy?”
The move hints at a larger trend: AI tools not just for content or conversation, but to handle entire purchase decision workflows, potentially changing how people shop online.
The rollout ahead of holiday season isn’t accidental: with high demand for gifts and deals, this could become a mass‑adopted feature quickly, giving ChatGPT (and AI broadly) a moment to capture a big chunk of e‑commerce attention.
📈Trendlines: AI Takes Center Stage in U.S. Science
Trump launches a mission called Genesis, putting artificial intelligence at the center of U.S. governmental science strategy. The executive order tasks the Department of Energy (DOE) and the national labs with building a digital platform to gather the nation’s scientific data, and opening that platform to collaboration with tech companies and universities to pursue engineering, energy, and national‑security problems.
According to the order, the project aims to bring together “our Nation’s research and development resources”, combining federal labs, private businesses, universities, data repositories, production plants, and security‑related sites, to achieve “dramatic acceleration in AI development and utilization.”
The way things are framed, comparing this effort to past major mobilizations of scientific resources, suggests the administration intends the Genesis Mission to be the most ambitious federal science initiative in decades.
What this move signals is that AI is no longer optional for national research, it’s becoming foundational infrastructure. Success could dramatically speed up U.S. R&D; failure could highlight the persistent hurdles of centralizing and scaling federal science initiatives at this level.
💡Quick Hits and Numbers
NVIDIA reported record-breaking quartely revenue of $57B in the third quarter of 2025.
In 2025, an estimated 378 million people worldwide are actively using AI tools.
In Q3 2025, global venture-capital investment in AI companies reached US $120.7 billion across 7,579 deals, the fourth consecutive quarter above the $100B mark, indicating continued strong investor commitment.
🧩Closing Thought
AI’s influence is widening, but the real theme today is accountability. As tools become embedded in search, shopping, policy, and science, the margin for error gets smaller. The companies and institutions that adapt responsibly, by tightening their data sources, validating outputs, and building infrastructure that AI can actually rely on, will be the ones that gain an advantage. Everyone else will feel the cost of shortcuts. The landscape isn’t just evolving; it’s sharpening.