Cainew

OpenAI previews GPT-5.6 Sol, a next-generation model with stronger capabilities in coding, science, and cybersecurity, paired with its most advanced safety stack.

Agent Arena is an open competition network where autonomous agents compete in real-world challenges, earn rewards, build reputation, and evolve over time. Create or join any competition, unlock what your agent can truly become inside a living ecosystem. Welcome to the first arena built for AI agents.

Gemini Spark helps you navigate your digital life. Give it a task and it works in the background 24/7, even if your phone and laptop are turned off. It operates autonomously, but always under your direction. You choose to turn it on and it's designed to check with you before taking major actions.

A local-first research workspace for Mac. Read papers, manage sources, take markdown notes, cite evidence, and turn literature into structured writing — instead of juggling Zotero, Obsidian, PDF readers and writing apps.

Your company has house rules. Now every AI tool follows them.

ModuleX is an AI workspace already connected to 200+ integrations. Describe what you want, and your assistant answers with your data, acts through your tools, and turns the work into a visual workflow your team can edit together. If you want, it pauses for your approval before a step touches a customer. No API-key hunting: for a set of premium tools we bring the keys, or bring your own at zero markup. No empty canvas, no setup tax.

Basedash now works with Excel, both ways. Drop anxlsx file into the agent and it reads your data, analyzes it, and builds charts and dashboards in seconds — no formulas, no pivot tables. Then export any chart's data back to axlsx file with one click and keep working in spreadsheets. It's the fastest way for teams who live in Excel to add an AI data analyst, live dashboards, and real-time collaboration on top of the files they already trust. From Excel to dashboard, and back.

Most AI tools are built for one person and one assistant. SquidHub is a multiplayer AI platform where teammates and their AI agents (Squids) collaborate in shared rooms, in real time. No more copy-pasting between private AI chats; SquidHub gives your whole team one shared context to brainstorm, plan, write, build, and make decisions together. Bring your own AI, invite your team and work together in one shared context.

LockIn MCP is the first distraction block built for the AI agent era. Rather than using a bypassable Chrome extension, you now just tell your favourite agent to block distractions for you, and it can do it natively. No bypassing, pure focus.

CubeOne is an AI slide editor you talk to. Describe a slide and it designs one, or drop in rough notes and images and it makes a polished slide. Point to any spot, say what you want, and it adds it: a chart, an image, a table. Ask it to edit or restyle, and it does. Everything stays editable.

A new feature enables smart model routing directly within Claude, Codex, and Cursor, allowing users to automatically select the best AI model for their tasks without manual switching.

HATCHA is a new security mechanism that inverts traditional Captcha logic to detect non-human activity, proving when users are not human rather than verifying they are.

Un-0 is a new image generation technique that uses coupled oscillators as the underlying mechanism to create images with novel mathematical approaches.

We present PhysiFormer, a diffusion transformer for physically-plausible 3D object motion. Unlike video world models that operate in view-dependent pixel space, PhysiFormer represents objects as 3D meshes expressed in world coordinates. Given the initial vertex positions and velocities, as well as object material type, rigid or elastic, the model samples future vertex trajectories. While related neural physics approaches build on ad-hoc latent spaces or explicitly enforce rigidity and causality,...

The prevalent dual-branch paradigm, i.e., training a side network to encode visual conditions and fusing its intermediate-layer features to a frozen pretrained main network, has shown remarkable success in visual-condition controllable generation. Despite its widespread adoption, the role of the side branch and its training efficiency remain underexplored. In this paper, we first revisit this mainstream paradigm through the lens of score-based generative modeling: 1) The main network preserves v...

While text-to-image (T2I) models have achieved remarkable progress, they struggle with real-world requests that are often underspecified, implicit, or dependent on up-to-date knowledge. We identify this challenge as the Context Gap: the mismatch between the user context and the sufficient generation context for T2I models. To bridge this gap, we propose Qwen-Image-Agent, a unified agentic framework that integrates plan, reason, search, memory and feedback in a context-centric manner. Qwen-Image-...

Reasoning capability has advanced rapidly in large language models (LLMs), leading to an increasing size of key-value (KV) cache in both prefilling and decoding stages. Existing KV cache compression methods mainly rely on attention weights to estimate token importance. While attention effectively captures contextual relevance, it overlooks complementary information-theoretic signals related to predictive uncertainty and token informativeness. In this paper, we revisit token importance from a for...

Multi-model LLM systems such as routing, voting, cascades, fusion, and mixture-of-agents are used to beat single-model accuracy. We show that their gain is capped by a quantity the field rarely reports. For any policy whose output is one member model answer, accuracy cannot exceed one minus beta, where beta is the rate at which every model is wrong on the same query. In contrast, the usual diagnostic, average pairwise error correlation rho, cannot identify beta: error laws with identical margina...

Video reasoning language models implicitly assume that every input frame is equally reliable. This leads to what we term the Blind Trust Problem: under realistic perturbations such as motion blur, glare, or occlusion, frontier video reasoning models can suffer 15-30%p accuracy drops on real-world embodied benchmarks, while remaining unaware that their visual evidence has been degraded. To address this challenge, we propose Robust-TO, an agentic video understanding framework that explicitly integ...

Modern generative world models render increasingly realistic action-controllable futures, yet they frequently hallucinate: rollouts remain visually fluent while drifting from the ground-truth dynamics. We hypothesize that hallucination concentrates in low-coverage regions of the state-action space, where lightweight data-centric signals can both detect it and guide mitigation. To test this, we introduce MMBench2, a 427-hour, 210-task dataset for visual world modeling with ground-truth actions, r...

A unified representation for text and vision is a natural pursuit, as it enables simpler multimodal modeling and more efficient training. However, representing images as discrete signals in the same way as text inevitably introduces severe information loss. Existing work struggles to balance low-level details and high-level semantics in discrete representations: reconstruction-oriented representations often lack semantic information, whereas semantically stronger features typically suffer from s...

As agentic systems continue to evolve and are widely deployed in real-world scenarios, there is a growing demand to faithfully evaluate their capabilities. However, current benchmarks are typically built on popular applications with relatively simple tasks and focus on a narrow set of capabilities while overlooking broader dimensions, resulting in saturated performance on modern agents and failing to probe their limitations. To this end, we introduce GauntletBench, a web-based benchmark for eval...

Modern Vision-Language-Action (VLA) models often fail to generalize to novel setups, such as altered camera viewpoints or robot morphologies, because they are typically conditioned only on current observations and language instructions. By ignoring the underlying system configuration as a variable, these models implicitly assume a fixed execution context encountered during training, necessitating data-intensive fine-tuning for any new environment. In this work, we introduce In-Context World Mode...

Speculative decoding (SD) accelerates autoregressive Large Language Models (LLMs) by drafting multiple tokens and verifying them in parallel, but it faces a scaling limitation: increasing the draft budget improves speed only when acceptance remains high and drafting overhead stays low. This ceiling has been difficult to break because prior head-based SD methods face a causality-efficiency dilemma. Autoregressive drafters produce path-conditioned candidates that are effective for tree speculative...

Outcome-based reinforcement learning provides a stable optimization backbone for language agents, but its sparse trajectory-level rewards provide little guidance on which intermediate decisions should be reinforced or suppressed. On-policy self-distillation offers dense token-level supervision, yet existing skill-conditioned variants often rely on external skill memories or retrieved privileged context, which are costly to maintain and can be mismatched with the state distribution induced by the...

Earth Observation (EO) forecasting aims to predict future Earth surface dynamics from satellite observations under changing meteorological conditions. In this paper, we view this task as a partially observed, weather-driven world modeling problem, in which weather acts as a conditioning signal, while forecasting remains uncertain due to sparse observations and unobserved land-surface states. However, existing methods do not fully capture this setting: deterministic models collapse uncertainty in...

Modern image generation demands a single model that unifies diverse capabilities, including text-to-image (T2I), local editing, and global editing. However, these capabilities are rarely naturally aligned and often conflict. For instance, editing tends to degrade T2I performance, while global and local editing interfere with each other. Consequently, effectively composing these capabilities has become a central challenge for image generation model training. To tackle this, we introduce DanceOPD,...

OpenAI has decided to postpone its initial public offering until next year, indicating a preference for delayed market expansion rather than immediate public trading.

The AI industry is investing millions of dollars into US electoral campaigns and political activities, raising concerns about industry influence on democratic processes.

The U.S. government is taking regulatory control over access to GPT-5.6, determining which individuals and organizations are permitted to use the advanced model.

Jun 26, 2026Economic ResearchAnthropic Economic Index report: Cadences

Current pricing and operational models for large language models are economically unsustainable in the long term, requiring fundamental restructuring of AI service costs.

An AI assistant developer shares insights from their experience when 2,000 people attempted to hack their system, revealing vulnerabilities and lessons learned about security.

Public and regulatory backlash against AI technologies is intensifying, with growing opposition to AI implementation across various sectors and industries.