In partnership with
☀️ TRENDING AI NEWS 🛠️ Agent SQL: NVIDIA released Auto Ontology, an Apache 2.0 semantic layer that grounds agent SQL in certified definitions. 🚨 Surveillance: A WA police trial scanned over 1 million faces, leading to 122 arrests and 9 false alerts. 🏢 Airlines: Ryanair is weighing pay freezes, fewer crew and more AI to offset a possible €2bn jump in its fuel bill. 🤖 Research: qwenGPU compiles a 4-lane SIMT GPU into Qwen3.5-2B as 45 untrained layers that run real CUDA. |
Twenty-five years ago, Tim Berners-Lee imagined an internet built for software agents to navigate on our behalf. Today, that idea stops being a thought experiment, and the plumbing for it is showing up everywhere you look.
| 🛠️ Tool of the day | |
n8n - a workflow automation platform that mixes AI capabilities with business process automation. It helps technical teams connect apps and build AI agent workflows with the flexibility of code and the speed of no-code.
| ⚡ Quick hits | |
Abu Dhabi and OpenAI signed a series of agreements, including a five-year deal between Inception42 and OpenAI to roll out frontier AI across government, enterprise and education. Read more →
South Korea's president proposed 21.3 trillion won of 2027 spending for a chip hub and data centers, physical AI and AI promotion. Read more →
iQIYI opened an AI Film Studio in Shanghai and began filming its first project, the internet feature Cop in Trouble. Read more →
| 🤓 AI Trivia | |
What did Tim Berners-Lee call his 2001 vision of an internet designed to be navigated by autonomous software agents?
The answer is hiding near the bottom of today's newsletter... keep scrolling. 👇
| 🤖 The Agentic Web Is Here | |
Berners-Lee's old idea gets its moment
A decade after he invented the World Wide Web, Berners-Lee co-authored a Scientific American article describing an internet that autonomous software agents could navigate to handle our everyday tasks. MIT Technology Review argues that vision is finally materializing.
The timing fits what we keep seeing across the industry: AI agents are moving from demos into real infrastructure, with standards, tooling and enterprise plumbing arriving all at once.
The bottom line
If agents become the primary way people use the web, the sites and tools that expose clean, machine-readable context will win. Design for the agent as well as the human visitor.
The Dell Pro with built-in intelligence.
Built for business. Ready for anything.
With built-in intelligence, you can stay productive and focused on growing your small business. So no matter what the day throws at you, your Dell Pro can handle it.
| 🛠️ NVIDIA Open-Sources a Semantic Layer for Agent SQL | |
Certified definitions, not guesswork
NVIDIA released Auto Ontology, an Apache 2.0 library that generates a semantic layer from enterprise metadata. The goal is simple: when an agent writes SQL, it should lean on certified business definitions instead of guessing what a column means.
The context is served through MCP, a CLI and an API, so it plugs into the agent stacks teams are already building. It is a quiet but practical fix for one of the biggest enterprise complaints: agents that confidently query the wrong table.
The bottom line
Open, certified definitions are the unglamorous layer that makes data agents trustworthy. If you run analytics agents in production, this is worth a look this week.
| ⚠️ One Million Faces Scanned in Western Australia | |
122 arrests, nine false alerts
A five-month trial of real-time facial recognition in Western Australia used surveillance vans to scan more than one million people. Police report 122 arrests and nine incorrect alerts.
Officers also engaged with 146 registered sex offenders during the deployments. Nine false alerts is a low rate on paper, but at this scale even small error rates land on real people.
The bottom line
The effectiveness numbers will be used to argue for wider rollout. The harder questions about consent, oversight and what happens to non-matching scans are still open.
| 🏢 Ryanair Eyes AI as Fuel Costs Bite | |
Pay freezes, fewer crew, more AI
Europe's largest low-cost carrier is weighing cost-cutting measures including pay freezes, reducing crew numbers on flights and increasing its use of AI. The trigger is soaring fuel costs tied to the Middle East conflict, with warnings of a roughly €2bn rise in its annual fuel bill.
It is a vivid example of AI entering the picture as a cost lever rather than a product feature, a pattern we track under job automation. The airline also expects European carriers to raise ticket prices.
The bottom line
When margins get squeezed, AI becomes the line item executives reach for. Expect more of this framing in earnings calls as costs climb.
| 🔬 qwenGPU Compiles a GPU Into Qwen3.5-2B | |
A model that runs CUDA by itself
qwenGPU compiles a 4-lane SIMT GPU into Alibaba's Qwen 3.5-2B as 45 untrained layers. The model writes every step of real CUDA launches itself, using Gated DeltaNet state as registers and memory.
It is a research curiosity more than a product, but it shows how far people are pushing the idea of weights as computation, not just as statistical pattern storage.
The bottom line
Nobody should swap their GPU for this. But it hints at how blurry the line between model and machine is getting.
| ⚠️ Chatbots Are Leading Hikers Astray | |
Outsourced judgment in the backcountry
Wired reports that outsourcing trip planning to AI can lead people to surrender their own critical thinking, with potentially deadly results in places like the backcountry.
The failure mode is familiar: a confident-sounding answer that is wrong, delivered to someone with no easy way to check it and no cell signal to double-check.
The bottom line
Use chatbots to brainstorm routes, never to verify conditions. Official maps, rangers and current trail reports still win.
Quick note: if today's stories sparked an idea for a project, our sponsor 60sec.site is an AI website builder that gets you a live page in about a minute. And for more daily AI coverage, visit dailyinference.com.
| 🌎 Trivia Reveal | |
The answer is the Semantic Web! Berners-Lee, along with two colleagues, laid it out in Scientific American in 2001, imagining autonomous agents that could navigate the internet for us. Today's agentic web is finally catching up to that vision.
| 💬 Quick Question | |
Would you trust an AI agent to run a task end to end for you, like booking travel or querying your company's data, or do you still want to approve every step? Hit reply and tell me where your line is. I read every response!
That's all for today - see you tomorrow with more!
| 🎁 Share Daily Inference | |
Refer a friend to unlock the AI Tools Starter Kit with 1 referral and the AI Insider Briefing with 5 referrals.