- AI Weekly Wrap-Up
- Posts
- New Post 11-19-2025
New Post 11-19-2025
Top Story
Google Gemini: One model to rule them all?
For much of the past year or so, the various AI model companies have been racing toward the future as a pack, with no one model clearly superior to all the others. That may have just changed. Google’s newest model, Gemini 3 Pro, has crushed all the usual benchmarks, showing superiority pretty much across the board over its closest rivals, Anthropic’s Claude, OpenAI’s ChatGPT, and X AI’s Grok. Whether this is a temporary victory or a prelude to a longer-lasting dominance remains to be seen.
The new model is being made immediately available in the Gemini phone app, as well as being embedded in all Google products, including its Chrome browser, and in its “AI Overview” search function.
Where we stop, nobody (yet) knows…
Clash of the Titans
Microsoft is embedding AI in all its products
Microsoft unveiled its two-fold strategy for the next phase of the AI wars: 1) embed AI into all of its current productivity apps (Word, Excel, PowerPoint, etc.), and 2) allow users to build their own workflows, agents and apps, using AI.
Venerable Excel will get turbocharged with AI that can write computer code and draw charts just by asking it. Word and PowerPoint will get AI upgrades that will make producing designs and text much easier.
Microsoft’s second strategic initiative aims to empower users to devise their own workflows, agents, and even apps on demand, without writing a single line of computer code. These creations will be tracked and managed by a single agent management system. allowing the company IT department to oversee resources and security.
Microsoft is very cleverly using its massive existing customer base to distribute its AI offerings. Users won’t ever have to choose to use AI - it will just be there embedded in familiar productivity products they already use.

Microsoft announces that it wants to be an AI-first company.
OpenAI pilots group chats in ChatGPT
As part of its ever-expanding quest to replicate all of the revenue-generating parts of the internet, OpenAI is testing group chats within ChatGPT. Users in Japan, South Korea, Taiwan, and New Zealand will be able to collaborate with one another,. and with ChatGPT, in a group chat within the ChatGPT app or website. ChatGPT acts as another participant in the group chat, and responds or stays quiet as the conversation seems to indicate. Members can always invoke ChatGPT explicitly by mentioning it in their post. In this way, ChatGPT can be used as a helpful assistant within the flow of a group discussion or project.

Planning a night out with friends? Add ChatGPT to your group chat.
Anthropic disrupts first reported AI-led cyberattack for espionage
Last week, Anthropic (maker of the Claude AI model) reported that they had detected and disrupted an espionage campaign that was almost entirely orchestrated by an AI agent, with only occasional interventions by the human perpetrators. Anthropic stated that they had noticed suspicious activity in mid-September, which investigation revealed to be a wide-ranging and sophisticated campaign of espionage against some 30 targets, including large tech companies, financial institutions, chemical manufacturing firms, and government agencies. The perpetrators were using Anthropic’s coding app known as Claude Code in their efforts to infiltrate these targets, and in a small number of cases, the attempts were successful. Anthropic banned accounts of the perpetrators as soon as they were identified, notified the targets, and coordinated with authorities. The perpetrators are believed to be part of a Chinese state-sponsored group. These events indicate how AI is heightening the issues with cyberattacks, and make a strong case for active use of AI for cyberdefense against such AI-orchestrated attacks.

Anthropic detected a cyberattack that tried to use its app for espionage.
Fun News
Bezos backs and heads AI startup for advanced manufacturing
Jeff Bezos has stepped back into a CEO role just 4 years after stepping down at Amazon - but this time he is heading up an AI startup that aims to remake manufacturing. The startup, known as Project Prometheus, is backed by $6.2 billion from Bezos and other investors. It aims to use AI to streamline engineering and manufacturing in the aerospace, automotive, and information technology sectors. Although the Prometheus Project is just out of stealth mode, it reportedly already has over 100 employees, poached from other AI labs such as OpenAI, Meta/Facebook, and Google’s DeepMind.

Bezos will be co-CEO of an AI startup for manufacturing.
This week’s #1 country song was generated by AI
The number 1 song in the US on Billboard’s “Country Digital Song Sales” chart is “Walk My Walk” by new star Breaking Rust. It turns out that Breaking Rust is an AI avatar, and the song was generated by AI. This is the first time that an AI-created song has topped the charts, and the reactions from industry veterans are decidedly mixed. Many artists are concerned that AI content will crowd out human content. In any event, Breaking Rust has released several other songs, which are also doing well.

Country music sensation Breaking Rust is an AI avatar, with AI-generated songs.
Teens disrupt pesticide industry with AI
Two teenage boys, Tyler Rose 18, and Navvye Anand 19, have founded Bindwell, a startup that uses AI to discover new pesticides that are more targeted and more ecologically safe than current alternatives. The pesticide industry is dominated by large firms that use chemicals that are often decades old. The teens saw the opportunity to innovate, and used the same AI tools that drug companies use for discovery of new medications, but applied the AI to finding new and better pesticides. Their business model is also clever - they will license their discoveries to the large pesticide manufacturers, rather than trying to compete with them head to head. The two teens have just raised $6 million in initial funding from topflight VC investors to launch their company.

Tyler and Navvye just raised $6 million for their AI startup. What did you do after high school?
Robots
Poop drones keep sewers running so humans don’t have to
Keeping America’s 1.3 million miles of sewers open and functioning requires regular inspections, which are expensive and time-consuming for humans to perform. Now a new breed of drone is being pressed into service, flying through the underground tunnels while recording a 3-D view of the walls, searching for breaks and cracks. Designed for operation inside dark and dank sewer pipes, they tend to have protective cages, high-intensity lights, high-resolution cameras, and various sensors to detect gas pockets and other environmental hazards. Drones can typically inspect over twice as many feet of sewer line per day as human teams, at a fraction of the cost, so business is booming in the poop-drone sector.

Swiss drone Elios 3 has a protective metal cage, lights, cameras, and sensors for sewer inspections.
Rover X1 is an affordable dog robot that can carry your groceries
Chinese robotics company Dobot is offering its semi-autonomous Rover X1 robot dog to consumers in the home market for a mere $1,000. With its sturdy industrial-grade body, cameras, and an AI brain for semi-autonomous functioning, the Rover X1 is being pitched for a variety of household uses - carrying groceries or other bulky items, patrolling the home and acting as a sentry, entertainment (it will perform tricks on command), and more. As China’s world-class manufacturing sector dramatically lowers costs of building robots, household and personal uses are likely to increase rapidly.

AI in Medicine
Mayo Clinic uses AI and genetics to identify rare heart condition
Researchers at the Mayo Clinic have used AI analysis of subtle patterns on a patient’s EKG to identify which patients will go on to develop a rare genetic disease called arrhythmogenic right ventricular cardiomyopathy, which can cause sudden death. The disease is linked to a mutation in the PKP2 gene, but not all patients with the mutation develop the disease. For those that do, the disease usually progresses silently until there is a fatal heart rhythm, or the heart is so weakened that the patient is debilitated, and may suffer an early death. By using AI to analyze the EKG’s of patients with PKP2 mutations, early signs of the disease can be detected, and preventive measures can be instituted. This should lead to fewer deaths, fewer poor outcomes, and less need for heart transplants.

ARVC gradually destroys the heart muscle of the right ventricle. AI can help prevent that.
Surgical robot used to perform first repair of brain aneurysms
US healthtech firm XCath has demonstrated that its EVR surgical robot is capable of repairing complex aneurysms in the human brain. The company’s robotic surgical system was successfully used on 3 patients in Panama recently, using standard stents and implants. Neurovascular repair is exceedingly delicate work, due to the fragility of both the blood vessels and the brain tissue. The company projects that its robotic surgery system will bring a level of precision to such procedures that will allow more surgeons to perform them, decreasing the current need for ultra-specialists.

These surgeons used a computer to guide the robot through delicate brain blood vessels.
That's a wrap! More news next week.