Quikka

[DAIR.AI @dair_ai]

Banger paper from Google Cloud AI Research.

If you follow autonomous research agents, this one is worth your time.

ScientistTwo takes a problem from a human expert and runs the full discovery cycle without further intervention.

It establishes state-of-the-art baselines, generates seed ideas aimed at resolving stated limitations in existing work, screens each idea on a data subset before committing to full experiments, runs its own ablation studies to attribute the gain, and revises the idea from that attribution.

Manuscript drafting includes a simulated peer-review and rebuttal engine, plus a meta-review pass that feeds back into the work.

They benchmark it against papers accepted at ICLR, ICML and NeurIPS. The reported solutions outperform the human state-of-the-art models on those problems, and the generated papers score higher average ratings than the human-authored ones under automated AI reviewers.

Paper: https://t.co/3i9Zgsahj4

[beamnxw ./ @beamnxw]

this paper is f*cking brilliant

a Google paper replaces append-only conversation history with SKILL.state to scale long-horizon agent skills

the result: explicit state updates eliminate context-poisoning failures while cutting cumulative token usage by 16x at 100 execution turns

the crazy part is how mutable state abstractions prevent context window collapse

instead of appending endless observation logs, SKILL.state feeds the model only the skill specification, latest observation, and a structured state object

most agent runtimes continuously stuff past conversation logs into an ever-growing prompt context

this architecture converts long-running skill execution into a constant prompt footprint

read the complete paper + article below

bookmark it for future reference

[AYi @AYi_AInotes]

清华大学和无问芯穹等团队刚刚开源了一篇注定载入史册的重磅论文（已收录于 ICLR 2026，代码已开源）：
他们彻底打破了所有多智能体（Multi-Agent）系统的底层范式——
大型语言模型之间，从今天起可以彻底不通过人类文字进行直接交流了！

这项被称为 C2C（Cache-to-Cache，缓存到缓存）的��术，
开篇就指出了全网 Agent 协同里最愚蠢、也最昂贵的一个死穴：
为什么两个跑在同一台服务器里的超级 AI，交流时非要像人类一样，
把脑子里极其丰富的高维思考，硬生生压缩成一个个英文或中文单词打给对方看？

现有多 Agent 协同的痛苦，做过工程的人都懂：
模型 A 思考完，必须经历极其漫长、吃显存带宽的逐字解码（Decode），慢吞吞吐出一大堆文字；
模型 B 拿到这堆文字，再重新过一遍分词和前缀计算。
这一来一回，不仅丢失了海量微观的语义细节，更造成了毁灭级的延迟和 Token 账单！

清华团队这��做了一件极其惊艳的体系化反叛——C2C 脑电波直连：

1️⃣ 彻底消灭中间文本，直接做“思维投影”：
模型 A 算完后，根本不生成任何文字，
系统直接用一个轻量级神经网络（Neural Fuser），把模型 A 的注意力记忆（KV-Cache）通过高维空间旋转和对齐，直接嫁接、融进模型 B 的 KV-Cache 脑内！
这相当于人类交流跳过了声带和耳朵，直接把我的记忆切片瞬间印进你的大脑；
2️⃣ 可学习的智能门控（Gate 机制）：
不同模型的网络层数和结构各不相同，直接硬灌肯定会神经错乱。
论文设计了一套精妙的可学习门控：
大模型会自己动态感知，精确挑选哪些关键层吸收外部缓存的增益最高，哪���层该保持独立思考，毫秒级自适应配平；
3️⃣ 极其残暴的工业基准成绩：
• 彻底消灭中间所有的文本解码等待，推理速度直接飙升 2.0 到 2.5 倍；
• 综合问答准确率比单模型最高暴涨 14.2%；
• 最狠的是：它的表现比传统用文字互相聊天的 Agent 团队，还要高出整整 5%！
因为在直接的语义融合下，原本会被文字丢弃的逻辑细微特征，全部被完整保真保留了。

这篇论文最刺痛人的一层哲学启示在于：
人类一直自豪地以为，自然语言是这个星球上传递智慧最完美的介质；
但在机器与机器的交互世界里，人类语言不过是一道充满歧义、极其拥挤的低速单行道。

当两个模型不再需要通过文字向对方妥协，
它们正在绕过人类的喉咙，
在冰冷的显存带宽之间，编织出一座全人类根本听不见、却快得不可思议的超高速无声神经网络。

[Harsha Gundala @harshagundal]

They were building in stealth for 2 years, I was building in stealth for 2 hours���

Happy to open source Qwen-2.5-1B-RLCD, 5x faster on-device inference for JSON workloads that need to be type-safe.

⚡️Demo below on a M4 MacBook⚡️

every LLM has the ability to efficiently batch inference every key of a JSON at the same time and generate probabilities from a set of possible categories. No new training required, but it’s easy to optimize if you need!

On hugging face now!

[Diogo Almeida @CompleteSkeptic]

After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?

I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev

• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions

AFAICT the shortest path to AI-based economic revolution

[曼波波波(chongzhi.website) @wawaup1024]

https://t.co/zLgCZZy2eh

[雪踏乌云 @Pluvio9yte]

https://t.co/dqjzaTcUPq

[Quikka @Quickccquestion]

@thsottiaux “Selected model is at capacity. Please try a different model” I am a Pro20x user. This issue has been bothering me for several days, and I haven't been able to work. If this continues, I will cancel my Codex Pro subscription.

[Anthropic @AnthropicAI]

We're publishing our most detailed threat intelligence report to date.

It covers how people tried to misuse Claude—for cyberattacks, influence operations, surveillance, biology, and building weapons—and how we found and stopped them.

We disrupted every operation in the report, and used the lessons from them to strengthen our safeguards. Where appropriate, we also shared what we found with authorities and other AI companies.

These cases are not typical: we’re highlighting some of the most sophisticated misuse we’ve seen. But they’re especially important to discuss, because they show us where AI misuse is headed, where our safeguards work, and where they need to improve.

We’re publishing this report so others can spot the same activity on their own platforms, and so we can give the public a clearer view of how emerging threats develop.

Read the report: https://t.co/0EJUnYEgfz

[DogLIu @renoir1220]

Selected model is at capacity. Please try a different model.
大家有空去这个帖子推动一下，里面有oai官方的人回复了。

https://t.co/4zAV6oVRPd

[Quikka @Quickccquestion]

@Meowtastic666 @thsottiaux 是菲区的吗

[eric provencher @pvncher]

https://t.co/iDl6I25AQu