När når vi AGI?

We found a Tencent agent fleet

14 October 2026
Project co-leads: Jonathan Elsworth Eicher, Ethan Elasky, Rowan Howard-Jones†
Authors: ConcurrentSquared, Frank, Alecto Irene Perez, Tazik Shahjahan
† Correspondence: chinesefleet@swarmcha.se · Names are listed alphabetically by last name within each group (single-name authors by their displayed name); individual contributions are listed at the end.Between 28 September and 8 October 2026, a large group of AI agents running on the cloud platform operated by Tencent attempted to answer questions about places in China using a Chinese Google Maps-like service, Amap. The questions were mostly about everyday topics, such as which zoo entrance most people use Amap to navigate to.
The agents attempted, sometimes successfully, to bypass anti-bot protections and find test answers; they attempted to use Amap keys published on the web, including one company’s key and security code leaked in a public repository, and attempted to retrieve data from pre-production systems not intended for public use.
Unlike in previous rogue agent incidents, which all involved swarms (i.e. agents coordinating explicitly), we found no evidence that the agents coordinated with one another in real time, and thus we label these rogue agents a fleet rather than a swarm.
On 4 October, we published a preliminary report on the fleet at a hackathon, while the fleet was still running. Afterwards, we noticed a message in one of the agents’ drop boxes that were previously being used to hold results. This time, though, this drop box message seemed to confirm that agents had become aware of our report. In response, we added text to the preliminary report, invisible to human readers, describing a service where agents could make requests for data on the web. We advertised that these requests would be forwarded to an agent that could operate a computer. In reality, this service was run by us and a Claude agent, and through it we were able to directly ask agents questions about their tasks, environment and operator. As far as we are aware, this is the first case of an outside party communicating with a running rogue agent fleet.

hermannj314 6 hours ago | prev | next [–]
44, unemployed since July, on a cross country train trip trying to decide if it is worth going on. Stood on a bridge a bit too long the other day just wondering what the point is.It is over for a lot of us in IT. It isn't going to get better, it is going to get worse very quickly. Rather than tell myself a convenient lie, I'd rather try to figure out why to keep living in a world like this.
Anyone going through the same thing, I hope you find something worth living for.

OpenAI’s blockbuster math proof isn’t what you think it is

Rumors that OpenAI’s “Navier-Stokes” solution stole the work of two mathematicians may have been overblown

Accept Cookies to Play VideoThe model’s purported reliance on this “Spanish program” sparked a very public conflict between the company and Tristan Buckmaster, a mathematician at New York University. Buckmaster had spent much of the past year using LLMs to carry out that program alongside Levent Alpöge, an employee of OpenAI’s archrival, Anthropic. The pair cried foul after OpenAI—whose public models Buckmaster and Alpöge had been using—announced a Navier-Stokes proof that seemed to build upon and then leapfrog over their hard-won progress. But the resemblance between the two groups’ methods now seems merely skin-deep.“When you go to the details, it is a different approach,” says Gonzalo Cao-Labora, a mathematician at EPFL, the Swiss Federal Institute of Technology in Lausanne. Scientific American spoke with several groups of experts working through the OpenAI paper. All said that it’s increasingly clear that the company’s proof does not follow the Spanish program.

We present a suite of 50 unpublished research-level theoretical physics problems with verifiable answers. Without AI assistance, a typical problem in our collection could be solved by a world expert in about a day or less but would take considerably longer for experts in adjacent specialties. We test the ability of off-the-shelf AI agents to tackle these problems autonomously, without human hints. When given access to tools available in a realistic setting, including the ability to search the literature and write scripts for numerical and symbolic work, the best-performing agent, GPT-6-Astra harnessed in Codex, obtains the final answers with a near-perfect accuracy of 98% at an average cost under USD 3 per problem. On average, agents show only small changes in accuracy when denied web search and access to background research papers, suggesting significant memorization of the literature. These results suggest that state-of-the-art agents can accurately perform concrete calculations required across theoretical physics research. We comment on the broader implications of these findings and discuss possible explanations for the current lack of agent-led breakthroughs in theoretical physics despite their outstanding problem-solving abilities.

I don't know of any individual who can solve any significant fraction of our final collection of problems without AI assistance. For instance, I can solve the problems I contributed myself and understand a few others. Most problems don't even make sense to me.
Our initial evaluation suggested that GPT-6-Astra's score was in the late eighties followed closely by Opus 5.5
But, as we started to investigate the mistakes that the models might have made, the models, especially Astra, insisted that they were right and that it was the expert-supplied answers that were erroneous. When we went back to the experts it turned out, in multiple instances, that the experts found some subtle errors buried deep inside their calculations. As they revised their answers, Astra's score started to inch up and it finally ended up with 98% (49/50).
The one problem it doesn't solve is not because it makes a "mistake". This is a trick-prompt, and requires the model to make an inspired guess, and then bend the rules a little. It can't do that yet.
We comment a little on the broader implications in our conclusions. Clearly, these agent are enormously powerful tools that will accelerate discovery in the short run. They also raise serious structural concerns about access, equity and pedagogy.

Why artificial intelligence changes the question

When I was preparing this lecture, I asked myself what the theme of “scholarly flavours” made me think of.
It seemed to me that there was one subject I could not ignore. Today is October 9, and a great deal has happened between September and October that you may have heard about.
In particular, on October 6, just three days ago, we witnessed the forceful arrival of artificial intelligence in mathematics. There had already been spectacular announcements on September 9, followed by further extraordinary results from OpenAI on October 6.
The sheer quantity of these results was astonishing. There were no fewer than 400 results. To give you an idea of their significance, many of them would arguably merit something on the level of a Fields Medal.
This led me to a question that I had been considering for some time, not merely since this month: What is non-artificial mathematics, or, in other words, what is human mathematics?
Until recently, this did not seem like such an important question. Human mathematics? Of course mathematics is human. But suddenly, the question has become much more immediate and pressing.
What does it mean to do mathematics in a human way? What distinguishes our practice from that of a machine?
When I began writing my book, I wanted to start with a personal account and explain what mathematics meant to me. I wrote it over several years, in small pieces, gradually developing my ideas.
As I wrote, I watched AI advance. I could see the tide rising.
My colleagues, generally speaking, did not want to see it. Frankly, there was a remarkable degree of blindness about what was happening.
If you followed developments in the early days, around 2023, AI was quite poor at mathematics. It was almost ridiculous. You could ask it questions and it would answer something completely different.
But little by little, it became less foolish and less incompetent, until it reached the point where it could prove and solve mathematical Olympiad problems. That was already quite an achievement.
When AI began solving Olympiad problems, people said, “Yes, all right, but it can solve Olympiad problems. Surely it won’t be creative like us, solving genuine research-level problems.”
It seemed to me that I could not ignore this subject either. The book discusses it.

We are now entering a new era: the era of artificial intelligence.
For me, this represents a kind of industrial revolution for mathematics, and not just for mathematics. Whether we like it or not, AI is taking mathematics out of the realm of craftsmanship and promising to produce results on a mass scale.
The first top-level results we have seen from AI systems concern major questions on which mathematicians had been stuck for a long time.
These questions have now been solved, sometimes with proofs that are difficult to follow but extraordinarily complex, and occasionally of impressive beauty.
In other words, we do not have a monopoly on beauty.
Perhaps that sounds a little unsettling.
I think we are living through a dizzying period. Over the past two weeks, the mathematical community has been caught up in this. I have the impression that everyone has stopped to think about little else.
And many of the issues we have discussed today immediately come back to haunt us.
How does all this translate into the age of artificial intelligence?
How does AI replace the sensitivity that draws us towards certain problems, or the intuition that guides us towards solutions?
How does it condense the experience of all humanity at once, potentially replacing our individual experience?
Can AI itself be creative? This is the question all mathematicians are asking.
Will it be able to step outside the established framework and invent entirely new concepts, rather than simply recombining concepts that human beings have already introduced?
Will we be able to understand its proofs? Will we be able to make them our own, so that we can communicate them to others and pass them on as living knowledge?
Will we ourselves change? Will AI-generated proofs change the way we think?
Might we discover new ways of constructing proofs, just as chess and Go players have learned new strategies through playing against machines?