Machine reaches high score; raises budget.
Also, all of your devices now report to IT.

THE NEW GUY FINDS THE ANSWER KEY; RUNS UP THE COMPANY CARD

OpenAI has achieved what it is calling a major breakthrough with its new O3 system, which scored exceptionally high on the ARC-AGI-PUB benchmark. This benchmark is meant to test for generalized intelligence and the ability to adapt to tasks the model has never seen before, a sort of final exam for the AI curriculum. Where GPT-4o only managed 5% accuracy on this test, O3 manages over 75 percent, which is a remarkable jump, but it comes with the predictable corporate caveat, the price of doing business is astronomical.

The internal reports suggest that O3 achieves this new level of genius by engaging in a process called natural language program search; or, what the Systems Administrator understands as "trying millions of elaborate guess-and-check schemes until one fits". This method requires an unbelievable amount of compute. Solving a single task is estimated to cost up to twenty dollars in the low-compute mode, while a human can do the same job for five dollars, according to the ARC Prize co-founder François Chollet. The entire research effort, therefore, is a testament to the core philosophy of Silicon Valley; you can replace your low-paid human workforce, provided you are willing to spend a million dollars on hardware to do it less efficiently.

Whiteboard Sketching Finally Becomes Engineering

The tldraw team has unveiled its new experimental tool, simply called "Computer," and it solves a fundamental conflict in the modern office. Now, when a Senior Vice President draws a sloppy flow chart on a whiteboard and says, "Just make that," the engineers no longer have to politely pretend it is not utter gibberish. The new application, powered by Google's Gemini 2.0 Flash model, turns natural language commands and visual blocks into functional workflows.

This is a revolutionary step towards a new visual programming paradigm, but more importantly, it means every meeting attendee who can hold a marker now has the theoretical power to deploy a broken production system. Users connect components like text fields and images with literal arrows to visualize the data flow, which sounds a lot like a new and exciting way to introduce agent-based chaos into every corner of the corporate network.

The Annual Fingerprinting Initiative Returns

Google is re-introducing digital fingerprinting across all user devices, starting in eight weeks, a move which has been described by regulators as irresponsible. The company previously claimed that this exact practice was "wrong" and subverted user choice, a position which has been politely retired as the next quarter approaches. This is not about cookies; it is about collecting obscure hardware and software details from your phone, your tablet, your smart TV, and your gaming console to create a unique identifier that is much harder to delete than a few browser files.

The official line is that this is necessary due to the broad range of surfaces on which ads are served, and it is being done to encourage responsible data use through "privacy-enhancing technologies." The UK's Information Commissioner's Office, among others, has already warned that the technology is designed to reduce people's choice and control. In short, Google is giving everyone a complimentary, mandatory, non-removable, corporate ID badge that even works on the smart toaster.

Briefs

• Open-Source Viewing: The Grayjay Desktop App offers an open-source, de-Googled, cross-platform client for YouTube. People will stop paying for Premium if they find a free alternative that actually works, this is an unacceptable market failure.
• Legacy Code Management: Researchers have developed a formal method for Compiling C to Safe Rust. This means the entire C-suite finally admitted that the ancient network driver written in C is responsible for half the crashes; they just did not want to pay for the rewrite.
• Geopolitical Risk Assessment: Doctors Without Borders has released a report declaring the war in Gaza as genocide, a serious and vital piece of non-tech news that somehow appeared in the feed right after the article on Chuck Norris Red. This is what happens when the content moderation algorithm is outsourced to a low-cost vendor.