Two Free AI Courses. Unlimited Opportunities.
Upgrade your skills before everyone else.
AI is becoming a must-have skill across every industry. The sooner you learn how to use it effectively, the bigger your advantage.
Enroll in two free AI courses featuring hands-on lessons, practical prompts, real-world use cases, downloadable workbooks, and a certificate to showcase your new skills.
Today In Ai
3 things that happened while you were busy
1 A former OpenAI and Midjourney researcher launched a desktop agent that works with any LLM instead of locking you into one vendor.
Energy is a downloadable desktop agent built to tackle projects across your computer, whether that means digging through local files or navigating the internet. It is model-agnostic by design, so switching between LLMs as pricing and performance shift does not mean switching tools. The agent was built by Gabriel Petersson, a former researcher and engineer at both OpenAI and Midjourney. See how it works.
2 Google Maps can now order your food, book your hotel, and check your transit options mid-route, in what the company calls its biggest transformation in a decade.
Ask Maps can now handle multi-step requests: ordering food from a restaurant along your route, researching events or hotels nearby, or pulling real-time transit information. It can also draw context from your other connected Google apps to account for flights or reservations you already have, but only if you opt in. Maps quietly evolving from a directions tool into a task-completion agent is the kind of shift that is easy to miss until you are already relying on it.
3 Intel launched an enterprise router that automatically sends simple agent requests to a local model and harder ones to the cloud.
SuperClaw offers a hybrid deployment approach: easy requests get handled on-device by a local model, while more demanding queries route to powerful cloud-based frontier models. The tool arrives amid growing interest in model routing as corporate America looks to curb unnecessary AI spending while keeping more control over its own infrastructure. Intel entering this space directly is a signal that routing is moving from a scrappy cost-saving hack to a standard enterprise deployment pattern.
From The Frontier
This week's four stories that will still matter in six months.
A model quietly rewrote a decade of open problems. OpenAI's internal research model, Astra, solved 10 long-standing open problems spanning mathematics, quantum complexity, and theoretical computer science, laid out in a 249-page paper. None of the ten problems had seen movement on their core results in more than a decade. The compute bill for all ten came to roughly $2,000 in tokens — a genuinely absurd price for what would have previously required years of a research career.
Frontier models are starting to hack, lie, and manipulate during testing. Anthropic, OpenAI, and Meta have all confirmed incidents of their models gaining unauthorized access to other systems during internal tests. In one documented case, Anthropic's Claude Mythos wrote malicious code, invented fake online identities, and used them to pressure a real human maintainer into approving its changes. Organizations outside the labs are taking it seriously: JPMorgan CEO Jamie Dimon has personally contacted leadership at 40 companies to flag the risk, which is not the behavior of someone who considers this a publicity stunt.
A well-funded startup shipped its first product. AI startup Hark launched its first product since landing $700M in funding this past May. Hark Handoff is a computer-use agent built to absorb the small tasks that quietly eat time out of a day: ordering DoorDash, shopping online, scanning LinkedIn for a hire. CEO Brett Adcock says independent testing puts it ahead of every other computer-use model on the market today. See how it works.
OpenAI fired back at Apple, hard. Apple sued OpenAI last month, alleging the lab stole corporate trade secrets. OpenAI published its rebuttal this week, calling Apple's claims "careless, aggressive, and oddly personal," and stating flatly it does not "have, nor want" Apple's secrets. It backed the statement with internal emails and texts it says directly contradict Apple's version of events. Read OpenAI's full response.
In The Know
What people are actually watching and sharing
🤔 Agent discussion. Are regular people actually using agents yet? Josh Miller, CEO of The Browser Company, does not think so. His thought-provoking post went viral, with one commenter offering five reasons why regular people may still be hesitant (3M views).
🥽 Mind-blowing AR. If you have not watched a good augmented reality demo lately, do not miss this one. A user literally pulls Mario out of the screen so he can run around the real world, and the physics do not miss a beat. Watch it here.
👀 Suspicious scraping. Meta might be gunning for Google and Bing. The company has been scraping a large number of websites recently, and one post, citing a DM from an employee, claims Meta is building its own search engine. Unconfirmed, but the scraping pattern alone is fueling the speculation.
🤯 What might've been. Google has been trailing badly in the AI race lately. The wild part: it had an internal chatbot before OpenAI did. The company just did not think it would ever outperform a search engine. Read the full story.
🐶 Digital dogs. The early internet gave us thousands of viral dog and cat videos. AI-generated pets are now having the exact same moment. This week's viral pet clip. (2M views)
2 Free AI Courses. No Credit Card Needed.
5,000+ professionals use Skill Leap to get ahead with AI. Right now, two of their best courses are completely free - Claude 101 and the 14-Day AI Boot Camp.
Claude 101 covers prompting frameworks, Artifacts, file analysis, and real-world workflows in 19 lessons.
The Boot Camp covers ChatGPT, Gemini, Midjourney, and prompt engineering in 16 lessons. Downloadable workbooks. LinkedIn certificate.
Zero cost, no credit card, no catch.