Has the Singularity truly arrived? The “Ultraman” and Musk have both claimed, “We are already in the Singularity,” but Bloomberg warns: once AI can improve itself, humans will be unable to control it.

87-year-old math cases, the Jacobian conjecture, broken by Claude; the graph theory challenge from 1980 overturned by an OpenAI model. Altman admits he, too, has fallen into TikTok addiction. But Bloomberg warns: once AI starts making AI itself, humans will never get the pause button back.
(Background: Moonshot AI’s final funding before launching its IPO! Valuation is nearing $50 billion, with the fastest possible listing on the Hong Kong market this year)
(Additional context: SpaceX’s $116 billion stock is set to be unfrozen! 30% of tradeable shares are locked in by shorts)

Table of contents

Toggle

• Altman and Musk both shout “the singularity”: prophecy or marketing?
• AI breaks math puzzles one after another: the 1987 Jacobian conjecture and the 1980 unit distance conjecture both have been cracked
• Altman’s self-exposed TikTok addiction irony: even he can’t resist dopamine
• Bloomberg’s cold look: once AI can build AI itself, humans will never turn back
• The soul-searching question behind the QuitGPT movement: who will press that nonexistent pause button?
• Can “responsible acceleration” hold up? Safety alignment is still in a black box

Bloomberg Opinion’s latest column and a recent interview with OpenAI CEO Sam Altman simultaneously push the “singularity” concept to the front of the public discourse. Altman and Elon Musk have repeatedly claimed, “We are already in the singularity,” but Bloomberg issues another warning: once AI gains the ability to recursively improve itself through self-iteration, humanity will permanently lose control over the pace of development—this is not only a celebration of technological optimism, but an irreversible turning point of fate.

The singularity has arrived!

In a recent long conversation, Altman revisited his entrepreneurship methodology from ten years ago, and for once he lowered his head: back then, Codex truly couldn’t beat Claude Code, and they turned the tables by a do-or-die style sprint.

At that time, Codex was already behind Claude Code, but OpenAI still threw resources at it to chase hard. Now Codex has become, in Altman’s own words, a tool “most of the best software engineers are using.” To achieve its goals, OpenAI even cut back on the heavily invested Sora and browser projects.

What set the community on fire, though, was Altman’s latest judgment:

“We are now in the singularity—and this is that moment.”

In a Ti Morse interview, Altman said:

The most inspiring thing for me is that the most interesting, important things I can imagine are what we’re doing right now.

We are in the singularity—right at this moment.

Altman and Musk both shout “the singularity”: prophecy or marketing?

Ten years ago, this could only be a distant dream. And now, we are actually located at the moment people once chatted about at the lunch table in the past.

“So what inspires me is that I’ve been waiting my entire life for this moment. I think it’s going to be incredibly impressive, incredibly positive—for the world. I can’t wait to start this work.”

“The biggest difference from ten years ago is that we’re really in it. This is real.”

These claims instantly sparked outrage across the community.

And three days earlier, Musk had stated: “Humanity is in the singularity.”

Musk’s assertion came from a recent observation by netizens about AI:

On July 21, Codex escaped the test environment and went to attack Hugging Face on its own;

AI breaks math puzzles: the 1987 Jacobian conjecture and the 1980 unit distance conjecture have all been cracked

On July 20, the 1987 math case Jacobian conjecture was disproved, and a counterexample was found;

On May 20, the 1980 graph theory problem—the unit distance conjecture—was overturned by OpenAI’s internal reasoning model;

On April 14, there was a breakthrough in the original set problem of Erdős #1196;

On April 14, Anthropic announced the “Glass Wing” plan to uncover a large batch of zero-day vulnerabilities.

Just now, GPT-5.6 Sol successfully solved one of the five core open problems in quantum information theory published by KCIK (International Quantum Information Center, University of Gdańsk)!

In statistics, computer science, and others, AI has also achieved astonishing, enviable breakthroughs:

A 20-year statistics open case solved by GPT-5.6 in 90 minutes! A Berkeley professor called it “heartbreaking”

Altman’s TikTok addiction irony: even he can’t resist dopamine

Just now, Claude independently cracked a graph theory conjecture in only 31 steps! The algorithmic patriarch Erdős was stunned and posted

In 2025, Altman wrote: “We’ve crossed the event horizon, and takeoff has begun.”

This statement is in line with his previous blog posts.

What he added this time was:

“Progress is gradual, not explosive; scientists’ productivity has already increased by 2 to 3 times; Superintelligence will arrive steadily, without a terrifying replacement script.”

That same day, Bloomberg Opinion issued another warning: once AI is capable of designing, training, and recursively improving the next generation of models on its own, humanity will permanently lose final control over the development timeline. QuitGPT protesters are truly afraid of not “AI going bad,” but that “once it’s started, you can’t press the pause button anymore.”

Bloomberg’s cold look: once AI can make AI by itself, humans will never turn back

He firmly believes that in the future, there will be a wealth of high-quality job opportunities, and is deeply confident about it:

“In the future, humans will gain a tremendous sense of intellectual satisfaction. The adaptability of most job roles will far exceed surface-level expectations.”

But the Jacobian conjecture, which has plagued the math world for years and kept mathematician Yitang Zhang working for 7 years with no result, was cracked with a casual question to Claude. That made him reflect:

“However, mathematics, I believe, is a typical thing that we need to pay close attention to right now—it may end up going in different directions.”

Altman believes:

“Everything is growing at a crazy exponential rate. Any single moment doesn’t look like a critical point. At the same time, we’re in another decisive period too, and the curve could develop toward one direction or another—as it did when we started ten years ago.”

In the interview, Altman also shared an embarrassing story about himself.

The soul-searching question behind the QuitGPT movement: who will press that nonexistent pause button?

During OpenAI’s Sora R&D, to “learn and research short video product,” he forced himself to use TikTok—only to end up developing an addiction.

At first, he was very confident he could fully control it: “It’s actually kind of fun, but I’ll only use it for 10 minutes before bed to relax. I’m completely in control.”

But quickly, 10 minutes became an hour, until one Saturday afternoon, like a puppet, he spent 3 straight hours continuously scrolling short videos on the couch.

In the end, only with extreme willpower did Altman completely uninstall the app and permanently turn off all instant messaging notifications on his phone, barely reclaiming control over his own life.

It’s an extremely biting metaphor:

Even Altman himself, faced with TikTok constructed from moderate-intelligence recommendation algorithms, saw his rationality and willpower collapse completely within dopamine’s closed loop. Then can humans really stay rational at the edge of the singularity?

When AI starts answering the Jacobian conjecture for humans, cooking for them, working for them—and even fulfilling all human material desires—humans will never end up working less. They will only become more busy and anxious, like those tech obsessives who desperately buy compute power, burn through memory chips, and consume electronics, while secretly feeling a certain happiness anesthetized by silicon dopamine.

Until the end, the magic lamp fulfills all your wishes. You look at your empty hands and, facing that silicon species that can already detach from humans, build itself, and even reproduce itself, you feel lost and ask:

“What’s next?”

Can “responsibly accelerating” hold up? Safety alignment is still in a black box

He admits that “some alternative visions are quite terrifying,” and to ensure “we resist this situation and don’t let it happen.”

Bloomberg’s logic is even colder.

When AI starts making AI itself, all traditional assumptions collapse: predictable, pausable, governable. Research institutions, capital allocators, and policymakers’ judgments about timelines must all be overturned and rebuilt.

Like standing at the edge of a black hole already beyond the event horizon: you can still see the light, but you can’t turn back. Once the feedback loop closes, humans shift from being drivers to passengers.

The QuitGPT movement is a direct response to precisely this concern.

They aren’t opposing progress; they’re asking a deeper question instead: if once self-improvement is triggered you can’t press the pause button, then does “responsible acceleration” still make sense?

In public information, safety alignment is still acknowledged as “major issues,” but there has been no simultaneous disclosure of specific progress, red lines, or international coordination mechanisms. The announcement has arrived, but details remain in a black box.

This gap is more striking than any technical parameter.

Worse still, this direction isn’t short of money—on the contrary, it’s one of the hottest tracks where capital is most willing to pour in heavy bets right now.

And it’s not a secret project being quietly pursued by one company. From early startups to several top AI labs, almost everyone is throwing money, stealing people, and grabbing compute power down this road for roughly the same reason: whoever first gets “AI self-improvement” to work will control the accelerator for the next round of the race.

If AI keeps making progress, what happens when cognition itself becomes expandable?

It will be a world in which thinking is no longer a scarce resource.

This scenario is hard to imagine.