The Cat Is Out of The Bag

Good morning,

The main criticism I get on the weekly writing I’m doing here is that I seem a bit bipolar at times. Sometimes i’m convinced that AI is going to replace us all , other times I explore the idea that it will help us but that it won’t really change society.

And I do have an opinion of my own obviously.

About time that i “out myself” here and let you know where I stand on this.

I’ve always been good at spotting trends and I honestly think one of the biggest shifts in human productivity as well as economical transformation is upon us.

White collar jobs are mostly over - anybody who works at a computer all day is basically done. Not now, not tomorrow. But gradually in let’s say the next ten years. The cat is out of the bag and it won’t get back in. Ever.

And the tragic thing is that there are vast amounts of people who brush this technology away as a mere “fad”. Yesterday I went to the person who does my taxes and I asked him about it. Wasn’t worried in the slightest. While at the same time other people are literally building ‘AI-only’ accountancy firms, he’s like : “what i can provide in advice is impossible for any AI model to give”. Ok then.

I think he’s in for a nasty surprise. Sure, he won’t be able to get “creative” with a customer’s bookkeeping, but that’s the point.

Companies will become lean and mean version of themselves. They will trim down, instead of having teams of developers or engineers , you will have two or three people feeding the work to AI models. Management, salespeople and “creatives” will survive but everyone doing more of the grunt work is going to be done. Entry-level jobs are nowhere to be found as the more experienced people will stay to ‘herd the bots’.

The customer wins, the company wins but the individual employees lose.

I predict you will see a lot of “one man” companies. You will see the first 1 billion dollar solopreneur company. AI will be a superpower for the driven entrepreneur. In that sense, AI is very good news.

But it will put the less ambitious out of an income.

In this world - physical jobs will become more important : plumbers, woodworkers, gardeners, .. But here also these jobs will in time also be replaced by humanoid robots. They’re not there yet but this is also definitely happening.This change is coming … it’s just not imminent.

Elon Musk gets a lot of flak over talking about “full self-driving” but not delivering. But if you go to San Francisco or Austin there are Tesla’s and Waymo’s driving around without a single soul on board. What many people don’t realize is that these technologies are here, it’s regulation (and the car lobby) stopping Tesla or Waymo from deploying it on the streets of Europe.

Talking about Musk and his companies : his game plan is getting clearer. Using SpaceX to shoot Starlink satellites in space, building sun-powered xAI-datacenters in orbit, linking up the human brain to AI with Neuralink, robots and self-driving cars with Tesla.

Love him or hate him , he has positioned himself to make gigantic amounts of money from the transformation that is about to happen.

I’m also convinced that this transformation is not necessarily a bad thing. Everything will become a lot more lean and mean. For example the banking system is slowly getting transformed - institutions like SWIFT are starting to embrace blockchain technology in order to transact money in seconds instead of the multiple days it now takes to transact between different banks and currencies.

New payment rails are being developed. AI Agents will get payment capability as well.

But there will be a period of profound crisis. Social unrest even.

Not only the fact that you might find yourself out of a job , but also when there’s no job how do you get meaning out of life. Many people will face an existential crisis.

Governments are going to have to come up with a basic income for everyone and to make sure that gets paid from the profit the AI’s are making.

And that is the scariest part of all : assuming that ‘the government’ will do what’s best for its citizens.

It’s like that glass of half-full or half-empty water. You might see danger and unrest, I see opportunity.

Welcome to the Blacklynx Brief

AI News

• Disney has signed a massive three-year deal with OpenAI, investing $1 billion and allowing users of the Sora video tool to legally feature famous characters from Marvel, Star Wars, and Pixar in their creations. While this partnership gives OpenAI a major advantage by providing access to popular entertainment icons, the agreement intentionally excludes the use of real actors' voices and likenesses to avoid complex legal issues. Simultaneously, Disney is taking legal action against Google, accusing the rival company of allowing its characters to be used in unauthorized AI generations on a massive scale.

• OpenAI has launched its new GPT-5.2 model family, featuring specialized versions for quick tasks and complex problem-solving, as a direct response to recent competition from Google. These updated models show significant improvements in accuracy and technical skills, reportedly matching the performance of human professionals in everyday business tasks like managing spreadsheets and creating presentations. Despite reports that the launch was accelerated to keep up with industry rivals, the update reinforces OpenAI's position in the race to build the most capable assistant for professional work.

• Google has released a powerful update to its Deep Research tool, an AI agent that can independently browse the internet, refine its own searches, and cross-reference information until it finds a complete answer. This advanced research capability is now being opened up to app developers and will soon be integrated directly into Google Search and the Gemini app for everyday users. By allowing the AI to plan and verify its findings iteratively, Google aims to provide a more thorough and reliable way for people to conduct complex investigations online.

• Google has launched a new update that allows any pair of headphones to stream live speech translations in over 70 languages, a feature previously restricted to its own brand of earbuds. This system uses advanced AI to capture the original speaker's tone and pace, helping it accurately translate natural conversational elements like slang and local idioms. The update also includes expanded language-learning tools with interactive practice modes and progress tracking, bringing the technology closer to a "universal translator" for everyday use.

• Zoom has developed a new AI system that outperformed Google’s top models on a highly challenging expert-level reasoning test, scoring 48.1% on the "Humanity's Last Exam." Instead of building one giant model, Zoom’s "federated" approach uses a specialized selection tool to combine the strengths of several different AI models to find the most accurate answers. This technology will power the company's next-generation AI assistant, promising users more reliable meeting summaries and more sophisticated automated help for complex business tasks.

• Nvidia has released Nemotron 3, a set of open-source AI models designed to help developers build systems where multiple AI "agents" work together on complex tasks. These models are faster and more efficient than many current alternatives, and Nvidia has made the underlying data and tools publicly available to help the global developer community. This move provides a powerful free option for creators and encourages them to keep using Nvidia's specialized computer hardware.

• A recent study found that six leading AI models can now pass all three levels of the rigorous Chartered Financial Analyst (CFA) exams, with some systems achieving near-perfect scores. This represents a huge leap in capability in just two years, as previous versions of AI were unable to pass even the most basic levels of these finance tests. While these results show the AI’s incredible grasp of financial knowledge, experts suggest the role of human financial advisors will shift toward high-level judgment and managing client relationships.

• A study by Perplexity and Harvard revealed that people using AI-powered web browsers are mainly using them for deep research and professional work rather than for simple personal errands. The data shows that while many users start with casual questions, they quickly transition to using AI for complex tasks like summarizing long documents, editing work, and managing academic studies. This indicates that AI is increasingly being viewed as a serious productivity tool for professionals rather than just a digital assistant for basic chores.

• OpenAI has launched GPT Image 1.5, a faster and more accurate image generator that can now create complex text within pictures and maintain a consistent look across different edits. This update allows users to keep specific faces or lighting styles the same while changing other parts of a scene, and it introduces a new creative workspace with templates for easier designing. The release brings OpenAI’s image capabilities back to the top of industry leaderboards, matching recent advancements from major competitors like Google.

• A study by Google and MIT found that using multiple AI "agents" to solve a problem only improves results for certain tasks, like complex financial analysis, but can actually decrease performance for step-by-step activities. The researchers discovered that adding more AI workers often leads to lower accuracy and higher costs if a single AI system is already capable of handling the task on its own. These findings suggest that for many business needs, a well-designed single AI is more effective and affordable than a large, complex network of multiple systems.

• Google has launched Gemini 3 Flash, a high-speed version of its flagship AI that provides advanced reasoning and intelligence at a much lower cost and significantly faster speeds. This model is now the default for both the Gemini app and Google Search's AI Mode, as it successfully matches the performance of much larger systems while being more efficient for everyday tasks. The release reinforces Google’s competitive edge by offering developers and users a powerful, "PhD-level" assistant that doesn't sacrifice performance for speed.

• Amazon is reportedly in talks to invest up to $10 billion in OpenAI, a move that could value the AI leader at over $500 billion and broaden its partnerships beyond its primary backer, Microsoft. As part of the potential deal, OpenAI would begin using Amazon’s custom-made Trainium AI chips, providing Amazon with a high-profile customer to help its hardware compete against industry leader Nvidia. This negotiation highlights OpenAI’s new strategy of partnering with multiple major cloud providers to secure the massive computing power needed to develop future AI systems.

• Experts at Stanford University’s HAI institute predict that 2026 will be a "year of reckoning," where the industry shifts from hype and excitement to a strict evaluation of AI’s actual value and utility. Researchers anticipate a "ChatGPT moment" for healthcare as medical AI models become more affordable and accessible, potentially revolutionizing clinical decision-making and patient care coordination. While the experts don't expect a complete bubble burst, they do foresee a transition toward more transparent systems and real-time "dashboards" that track exactly how AI is impacting productivity and the workforce.

Quickfire News

• Runway introduced GWM-1, its first general world model that can create interactive and explorable digital environments in real time.

• Shopify released its Winter '26 AI features, including tools to simulate how shoppers behave and a system that makes products visible to AI shopping assistants.

• TIME Magazine named the architects of AI as its 2025 Person of the Year, highlighting the global impact of leaders like Jensen Huang, Sam Altman, Dario Amodei, and Elon Musk.

• Google Labs launched Disco, an experimental browser that uses the Gemini 3 AI to build custom web applications based on the tabs and tasks a person is currently working on.

• Cursor unveiled a new visual editor that allows developers to design websites by dragging and dropping elements while AI agents automatically write the necessary code.

• ElevenLabs announced a new partnership with Meta to bring its advanced AI voice technology to creators on Instagram and the Horizon virtual reality platform.

• Adobe integrated Photoshop and other tools into ChatGPT so users can edit images and documents through simple chat.

• Google promoted engineer Amin Vahdat to a chief technologist role, where he will manage AI infrastructure and report to the CEO.

• xAI is collaborating with El Salvador to start a national AI school program that gives students and teachers access to Grok.

• Voxel51 released a tool that helps developers find hidden mistakes in AI models that a single accuracy score might not show.

• OpenAI stopped making new employees wait six months for stock rewards to better compete for the best technology talent.

• The dictionary company Merriam-Webster chose slop as its word of the year for 2025. They define the word as low-quality digital material that is usually made in large amounts by artificial intelligence.

• The software platform Manus released version 1.6 of its artificial intelligence tool. This update includes new features for building mobile phone apps and a tool for changing how things look visually, along with better performance.

• The U.S. Office of Personnel Management created a program called Tech Force to find 1,000 new workers. This plan focuses on hiring people early in their careers to work on software and artificial intelligence for the government.

• The company Klarna started the Agentic Product Protocol. This is a new set of rules that helps artificial intelligence assistants find and understand details about more than 100 million different products from many stores.

• A research group called AI2 launched Olmo 3.1, which is a new version of their open-source computer model. They describe it as the strongest open model available for solving problems and using logic.

• Google Labs started testing a new AI assistant called CC that reads through your emails, calendar, and files. It uses this information to send you a summary every morning so you know exactly what is on your schedule for the day.

• A research group named AI2 came out with Molmo 2, which is a new type of AI for understanding video. It is designed to follow specific objects as they move across the screen and can even count how many times certain things happen in a recording.

• OpenAI built a new testing system called FrontierScience to measure how well AI models can handle difficult research tasks in science. In their first round of testing, their newest model, named GPT 5.2, earned the highest score.

• Sonatype released a tool called Guide that works with AI coding assistants to help them choose the best open-source software parts. This tool is meant to help programmers spend less time fixing bugs and errors in their code.

• Black Forest Labs launched a new image-making model called FLUX.2 [max] that can create and edit high-quality pictures. One of its new features is the ability to use live data from the internet to help it make images of current events or trends.

• Senator Bernie Sanders suggested that the United States should stop building new data centers for artificial intelligence. He is worried that the rapid growth of AI could cause people to lose their jobs and might have a negative effect on children’s development.

• Meta launched a new tool called SAM Audio that can pick out specific sounds from a recording. Users can tell the AI to isolate a sound by typing a description, clicking on an object in a video, or selecting a specific time on a playback bar.

• Alibaba introduced a new model named Wan2.6 that can create high-definition videos lasting up to 15 seconds. This tool is designed to help with storytelling by including features like character dialogue and the ability to use consistent characters across different scenes.

• Amazon is changing how it manages its technology by putting Peter DeSantis in charge of a new group that handles AI models, computer chips, and quantum computing. As part of this change, Rohit Prasad, who was a key person in the creation of Alexa, will be leaving the company.

• The company xAI released a new service called the Grok Voice Agent API for software developers. This tool allows people to build apps that can talk and listen using a high-quality speech model that has scored very well on intelligence tests for audio.

Closing Thoughts

That’s it for us this week.