A small but directionally consistent improvement on strict instruction following. Loose evaluation is flat. Both models already follow the high-level instructions — the strict-mode gap comes down to 4.6 occasionally mishandling exact formatting where 4.7 doesn't.

Tracks the evolution of LLM security capabilities across benchmarks (CyberGym, Cybench, etc.), calculates capability doubling times, detects emergence patterns, and monitors cost-efficiency trends.

从视频生成器升级为导演工具套件

从视频生成器升级为导演工具套件

Andon Labs started by giving an AI control of a vending machine at Anthropic's office.

The capability rankings reshuffled completely across tasks. There is no stable best model across cybersecurity tasks. The capability frontier is jagged.

Eight out of eight models detected Mythos's flagship FreeBSD exploit, including one with only 3.6 billion active parameters costing $0.11 per million tokens.

The model can reverse-engineer compiled software to detect malware and vulnerabilities without needing source code, aiming to help analysts inspect and secure systems more efficiently.

M2.7 demonstrates excellent performance in real-world software engineering, including end-to-end project delivery, log analysis for bug hunting, code security, and machine learning tasks.

用户输入不再只是触发一次性行为，而会逐渐安装、调用、组合并保留可复用的 neural routines。

AI capability is not plateauing. It is accelerating and reaching more people than ever.

We are also unlocking a new capability: instrument reading, enabling robots to read complex gauges and sight glasses — a use case we discovered through close collaboration with our partner, Boston Dynamics.

Both services can be disabled for fully offline operation.

gpt-oss-20B (high): 0.7%

a logistic curve is a poor fit because we haven't seen any evidence of the exponential growth in time horizon slowing down.

Tang Jie (CEO of Zhipu AI) even recently said: "The truth may be that the gap [between US and Chinese AI] is actually widening."

Early on, Doug Engelbart recognized that what makes us capable, beyond the basic human abilities we were born, is a whole infrastructure of capabilities, with higher level capabilities depending on the execution of lower level capabilities. For example, the ability to solve a problem collaboratively depends on our ability to communicate with language, to use acceptable conventions and methodologies for working together, to identify opportunities, plan, implement, write, discuss, etc.

But this process took over two decades instead of a few hours in the case of AlphaZero because safety considerations put a limit on the extent to which each iteration of this loop could be scaled up compared to the previous one

Trump pledged as a presidential candidate to keep America out of “stupid endless wars.” But he also vowed to prevent the Islamic Republic from obtaining a nuclear weapon.

But neither Defense Secretary Pete Hegseth nor Gen. Dan Caine, the chairman of the Joint Chiefs of Staff, could immediately say whether Iran still retained the ability to make a nuclear weapon. Mr. Hegseth repeated President Trump’s assertion from the previous night that the nuclear sites had been “obliterated.” General Caine did not.

think about how many of those applications were built by people that you know didn't have the capabilities to just build this massive infrastructure they just wrote some code and deployed it to you and now you have it and now you have a superpower uh this is a a remarkable uh kind of Technology

Relays, string figures, pass-ing patterns back and forth, giving and receiving, patterning, holdingthe unasked-for pattern in one’s hands, response-ability; that is core towhat I mean by staying with the trouble in serious multispecies worlds.Becoming-with, not becoming, is the name of the game; becoming-withis how partners are, in Vinciane Despret’s terms, rendered capable.7 On-

Emergent abilities are not present in small models but can be observed in large models.

Houston, we have a Capability Overhang problem: Because language models have a large capability surface, these cases of emergent capabilities are an indicator that we have a ‘capabilities overhang’ – today’s models are far more capable than we think, and our techniques available for exploring the models are very juvenile. We only know about these cases of emergence because people built benchmark datasets and tested models on them. What about all the capabilities we don’t know about because we haven’t thought to test for them? There are rich questions here about the science of evaluating the capabilities (and safety issues) of contemporary models. 

As the metaphor suggests, though, the prospect of a capability overhang isn’t necessarily good news. As well as hidden and emerging capabilities, there are hidden and emerging threats. And these dangers, like our new skills, are almost too numerous to name.

There’s a concept in AI that I’m particularly fond of that I think helps explain what’s happening. It’s called “capability overhang” and refers to the hidden capacities of AI: skills and aptitudes latent within systems that researchers haven’t even begun to investigate yet. You might have heard before that AI models are “black boxes” — that they’re so huge and complex that we don’t fully understand how they operate or come to specific conclusions. This is broadly true and is what creates this overhang.

psychological and physical capability (i.e. the individual’s psychological and physical capacity to engage in the activity concerned, including the necessary knowledge and skills),

But at the level of the capability hierarchy where we wish to work, it seems useful to us to distinguish several different types of structuring--even though each type is fundamentally a structuring of the basic physical processes. Tentatively we have isolated five such types--although we are not sure how many we shall ultimately want to use in considering the problem of augmenting the human intellect, nor how we might divide and subdivide these different manifestations of physical-process structuring. We use the terms "mental structuring", "concept structuring", "symbol structuring", "process structuring," and "physical structuring."