OpenAI

A SMALL KEYBOARD WITH BIG AMBITIONS

OpenAI has finally taken a step that many people have been expecting for a long time: it has released its own device. The surprise, however, is that it is neither a smart speaker nor a wearable gadget with a voice assistant. It is simply a keyboard—a small, compact keyboard equipped with a joystick and a rotary dial. Its name is Codex Micro.

The device was created in partnership with Work Louder, a Canadian-Italian company known for producing mechanical keyboards for technology enthusiasts and professionals. Visually, Codex Micro resembles some of the manufacturer’s other products, such as the Creator Micro 2, which was developed in collaboration with Figma. Internally, however, it is a completely different story.

The device looks like a compact control panel featuring thirteen illuminated keys, a mini joystick, and a rotary controller. All these elements are connected to Codex, OpenAI’s proprietary AI platform for programming. The idea is simple but ambitious: instead of clicking a mouse and navigating menus on a screen, developers can control their AI agents with the press of a button.

The LED keys display the status of individual agents: white means idle, blue indicates that a task is being processed, green means the task has been completed, and red signals an error. A single click allows the user to switch between different AI assistants, while a double-click brings the selected assistant to the foreground.

The joystick can be used to launch common tasks such as reviewing code, searching for errors, or performing refactoring. The rotary dial is the most interesting element of the device. It controls the AI’s “reasoning level”—in other words, how much time and computing power the neural network will spend solving a particular task. Turn it in one direction to receive a quick response to...

The most dangerous AI model finally enters the public sphere — under supervision

Seventy-two days. That is exactly how long the technology world spent in anxious anticipation.

On April 7, 2026, Anthropic quietly, almost unnoticed, released what many experts called “the most dangerous artificial intelligence model ever created.” Mythos.

A model capable of finding vulnerabilities in software. A model that can break through defenses programmers spent years building. A model that, if it falls into the wrong hands, could potentially disrupt banking systems, paralyze power grids, halt transportation networks, and cause widespread digital chaos.

Back in April, however, Anthropic acted cautiously. Instead of making Mythos publicly available—as some publicity-hungry startups might have done—the company granted access to only fifty organizations. These included operators of critical infrastructure: banks, energy providers, transportation networks, and government agencies.

Even then, access came with strict controls, usage restrictions, and monitoring of every request.

Now comes the next step.

On June 10, 2026, Anthropic introduced Claude Fable 5—the first Mythos-class model available to the broader public.

Well, “public” is a relative term.

The model is accessible through a paid API, subject to strict limitations, safety filters, and automatic redirection of dangerous requests to a less capable model, Claude Opus 4.8. Nevertheless, for the first time, ordinary developers, companies, and researchers can experiment with technology that was under lock and key only two months ago.

It is as if the U.S. government suddenly declassified all UFO files, or CERN opened the Large Hadron Collider to anyone who wanted to use it.

The technology is powerful enough to require extraordinary caution—and important enough that keeping it hidden indefinitely was no longer an option.

So what exactly is Mythos? Why is Claude Fable 5 considered dangerous? And why did Anthropic decide to release it despite the risks?

Mythos: What...

Promises That Haven’t Materialized

The world of artificial intelligence has grown accustomed to a relentless pace. Every week brings a new breakthrough. Every month sets a new benchmark. Companies compete to launch smarter models, more convenient APIs, and cheaper tokens. In this race, Meta Platforms has occupied the position of a confident middle-ground player—not the fastest, but not the slowest either.

Until now.

According to information published by The Wall Street Journal on Wednesday, Meta has repeatedly postponed the release of its latest AI model for developers. The delay has stretched to nearly two months, and more importantly, the company still has not committed to a new release date.

What is behind this delay? Why has a company that recently promoted its openness and rapid development suddenly gone quiet and started pushing back deadlines?

As is often the case in technology, the answer is more complicated than it appears. This is not simply a matter of unfinished code. It is a strategic pause—perhaps even a reassessment of the company’s entire approach.

The story began in April, when Meta unveiled its newest AI model, Muse Spark, with considerable fanfare. The company promised a model capable of competing with OpenAI’s GPT-4 and Google’s Gemini while remaining partially open, consistent with Meta’s broader philosophy. Developers around the world eagerly anticipated the launch of an API that would allow them to integrate Muse Spark’s capabilities into their own applications. The expected timeline was ambitious: the API would be released alongside the model itself.

April came and went. Then May.

The API never arrived.

A Promise Left Unfulfilled

According to sources cited by The Wall Street Journal, the head of Meta’s AI division assured developers that the release was coming “soon.” That statement was made nearly two months ago.

“Soon” has stretched considerably.

Developers who...

Anthropic and OpenAI have done what many had long expected but what participants in the shadow market refused to believe until the very last moment. Both companies updated their internal policies almost in sync, putting a firm stop to secondary trading of their shares. The wording published on their respective pages sounds nearly identical, as if the lawyers of two fierce rivals had been copying off each other. But that's not really the point. The point is that the two hottest AI startups on the planet have just cut off the oxygen supply to an entire industry that grew up around investors desperate to grab a piece of their businesses before either goes public.

At Anthropic, the language is brutally clear: any sale or transfer of securities without board approval is declared void. A buyer who risks entering such a transaction will not be recognized as a shareholder and will receive absolutely no rights whatsoever. OpenAI mirrors the exact same structure: without the company's written consent, any transfer of shares has neither legal nor economic value. In plain terms: if you bought without asking, you might as well have thrown your money to the wind.

An Entire List of Grey Schemes Now Banned

What's most interesting is that neither company limited itself to a generic prohibition. They named specific avenues that are now considered illegitimate. The list includes direct sales, special purpose vehicles, tokenized equity interests, and forward contracts. This is not a random collection of words. These are precisely the tools that the market has spent the last several years using to carve out detours toward shares in private AI giants.

The notorious SPVs — special purpose vehicles — deserve particular attention. The scheme is simple and elegant in its audacity. A shell company is created with the sole...