Cainew

Leanstral 1.5 has been released with performance improvements and new features for developers and organizations. The update continues the evolution of the platform toward more efficient and capable operations.

TabFM is a new zero-shot foundation model specifically designed to handle tabular data without requiring task-specific training. This approach enables practitioners to apply powerful AI capabilities directly to structured data problems.

Type what you need. Hold Acti Bar. Acti understands your intent and brings back the right result, link, or action - right where you are. Use Acti for live sports schedules, nearby restaurants, Notion docs, LinkedIn profiles, Meet links, Calendar actions, and custom workflows - without leaving the conversation.

Today's models are capable enough. Smart enough. Fast enough. But we still feel they don’t fit in the room. Humalike is building the behavioral infrastructure for humanlike AI agents. The social skills & proactiveness your agents have been missing. APIs, models, benchmarks.

Give /automate a task in plain English and it drives a real browser to do it: navigate a site, click through a multi-step flow, fill a form, reach a page that only renders after interaction. The result streams back in one API call. It's an API you call, not a framework you install. Browser and LLM included, nothing to host, no concurrency ceiling. Accessibility-tree automation spends 60 to 80% fewer tokens than screenshot-based agents. Built by Mozilla. Ephemeral, no training on your data.

Our most agentic Sonnet yet, with top-tier intelligence for coding and everyday professional work.

Adam brings AI CAD assistance into the tools mechanical engineers already use. Create & edit parts with prompts, reference selected geometry, clean up feature trees, and keep everything editable. All natively inside Onshape and Autodesk Fusion.

Email deliverability on autopilot. A subscription-based, AI-native email deliverability service where AI agents and human experts work as an extension of your team to audit your infrastructure, fix issues, monitor email health daily, and improve inbox placement with weekly reporting.

Sequence is the financial execution layer for AI agents. Unlike read-only tools, your agent uses the Sequence API to send, split, and route real money across all your bank accounts, cards, apps, and loans. Scoped API keys mean agents never hold your credentials, server-side spending limits keep you in control, and full audit trails log every action. The infrastructure is battle-tested in production on regulated rails, moving north of $3B. One-call integration from Claude, n8n, Zapier etc.

Give Mark your website and it researches your business, creates a personalized GTM plan, and builds web agents that automate lead gen, enrichment, outbound, SEO, and Google Ads campaigns. Using Mark is like vibe coding, but for sales and marketing campaigns. Built on Airtop's Agent Builder platform, Mark compiles every automation into deterministic code, so its agents run reliably and 10-100X cheaper than LLM-per-step agents. Real marketing automations, built just by typing.

Claude Science is your AI workbench for scientific research. Works through your research like a skilled scientist, running the analysis and tracing every step. Spend less time stitching pipelines together, and more time on the science.

You can now build apps directly in Fuser—no code required. Fuser is an end-to-end creative engine. Make images, video, sound, 3D and more on canvas. Your references, generations, media and data can become something your app reads, responds to, and evolves from. Prompt, generate, edit, and ship in one click. Go live in minutes. Try it now: app generations are free for the next month on fuser.studio 💙

Gemini Omni Flash (gemini-omni-flash-preview) just rolled out to developers via the Gemini API and Google AI Studio, natively supporting high-quality video generation and conversational editing from a combination of text, image and video inputs. This model is priced competitively at $0.10 per second of video output, which is the same as Veo 3.1 Fast.

OASIS Ring combines our patented ring trackpad with private voice capture, integrated with Wispr Flow. After shipping the most advanced trackpad ever built into a ring this year, we’re launching our next step: an interface on your finger that lets you whisper privately, dictate with Wispr, and edit text with our trackpad without ever touching a keyboard.

Tell RunInfra what you need and it builds the production API. No dashboards. No config. Describe any open source model or full app in plain language. We optimize it for real: benchmark GPUs, quantize the model, generate custom CUDA kernels with our Forge agent. It runs faster and cheaper than standard hosting. Build voice (speech → AI → speech), doc search, vision, or model routing, all in one chat. Pay per million tokens. Scale to zero. Run managed or on your own GPUs.

Stigg is the usage runtime for AI products: the real-time enforcement and governance layer between your app and your billing stack. It decides what every customer, user, team, and agent can do, the moment they try. Millisecond credit checks, zero overdraft, enterprise governance, and modular BYOC. Metering, credits, entitlements, and governance in one runtime. Enforce in the request path instead of reconciling on the invoice. Free forever for AI startups.

For knowledge workers orchestrating a dozen AI agents. 🧠 Your context never sits still: decisions shift, deals move, priorities change by the hour. Yet every new chat starts blank, and no agent knows what the others already worked out. N71 gives them one shared context that stays current, connect your tools and it maintains a living knowledge graph they read from over MCP, updated the moment anything changes. Ask any agent anything, and it's already caught up. 🔗

Researchers have demonstrated that matrix orthogonalization techniques can substantially improve memory capacity and performance in recurrent neural network models. This optimization approach offers practical benefits for sequence processing tasks.

Text-rich image generation is one of the most challenging settings in image generation, since models must simultaneously produce visually realistic images and render legible, semantically aligned, and layout-consistent text. Existing data pipelines usually follow a static crawl-filter-freeze paradigm. They collect candidate samples, filter them once, and freeze the accepted data for training. However, rejected samples are usually discarded, although they often contain useful failure signals such...

LLM agents increasingly act over long horizons, where a single trajectory can contain hundreds or thousands of actions. In these settings, outcome-only rewards provide too sparse guidance, failing to inform the model about the goodness of intermediate actions. Dense supervision methods aim to solve this problem by scoring intermediate steps, from intrinsic confidence to self-distillation and embedding similarities. However, it is common practice to evaluate them by measuring the downstream perfo...

Video World Models are interactive video generation models that predict future world states based on user actions and history video frames. A critical challenge in video world models is the lack of memory, causing inconsistent generated scenes over extended durations. Previous methods explored rule-based context frame retrieval as memory, but they fail to generalize in scenarios with scene occlusions and dynamic objects. We propose MemLearner, a learning-based adaptive context query method using...

Spoken language models (SLMs) extend LLMs to speech input and output. Existing SLMs represent speech at fixed frame rates (e.g., 25 or 12.5 Hz), ignoring the time-varying information density of speech and offering no flexibility to trade off quality for speed at inference time. Recent audio tokenizer research has proposed dynamic frame rate speech coding, which exploits this non-uniformity and enables two new capabilities: very low average frame rates and frame rate controllability. However, thi...

Graphical user interface (GUI) agents build on vision-language models to complete user tasks end-to-end in real applications through interface actions such as tapping, swiping, text entry, and navigation. However, existing GUI agents are trained and evaluated largely on offline trajectories, simulated environments, and standardized benchmarks. These differ substantially from real applications in interface layout, interaction logic, and abnormal-state distribution, and cannot faithfully character...

Foundation models have transformed vision and language processing by providing rich, reusable representations that transfer across diverse tasks. Sheet music, as a visual encoding of musical language, lacks such a strong domain-specific backbone. We introduce MuSViT (Music Score Vision Transformer): the first foundation vision model for sheet music representation -- a ViT encoder pre-trained via Masked Autoencoders on 9.7 million pages from the IMSLP. To handle the complexity of real-world score...

Metacognition is a critical component of intelligence that describes the ability to monitor and regulate one's own cognitive processes. Yet LLMs exhibit systemic deficiencies in key metacognitive faculties: they hallucinate with high confidence, fail to recognize knowledge boundaries, and misrepresent their internal uncertainty--undermining trustworthiness and reliability. Since monitoring task performance and adapting behavior accordingly are central to metacognition, we posit that models capab...

Agentic reinforcement learning requires assigning credit to environment-facing actions such as searches, clicks, edits, navigation commands, and object interactions. Standard GRPO uses the final verifier outcome as a uniform advantage over all action tokens. This outcome signal is useful but structurally incomplete: it punishes useful exploration in failed rollouts and reinforces redundant or regressive actions in successful rollouts. We propose TRIAGE, a role-typed credit assignment framework t...

Existing instruction-based video editing datasets commonly focus on single-task appearance editing, failing to meet the complex creative demands of real-world scenarios. To bridge this gap, we present Goku, a large-scale dataset featuring 2 million high-quality, instruction-aligned video editing pairs, which is the first to extend task boundaries from basic appearance editing to multi-task and structural manipulations(e.g., precise control of subject movement). To tackle the data synthesis chall...

Generative models have achieved remarkable progress, yet applying them to satellite imagery remains challenging. Unlike natural imagery, satellite scenes are structured by spatially complex and semantically distinct geometries. Prior work addresses this complexity by adapting natural image frameworks using dense rasters or sparse prompts, trading off annotation cost and fidelity while breaking compatibility with vector primitives commonly used to represent geographic information. We introduce Te...

Visual generative models are typically trained in two stages. A tokenizer is first trained for reconstruction and then frozen, after which a generator is trained on its discrete indices or continuous latents. This decoupling leaves the tokenizer unaware of what the generator finds easy to model. We present GEAR (Guided End-to-end AutoRegression), which trains a vector-quantized (VQ) tokenizer and an autoregressive (AR) generator jointly and end-to-end, guided by representation alignment. The key...

Block Diffusion Language Models (BD-LMs) improve diffusion-based text generation with KV caching and flexible-length generation. A natural next step is to extend them from Single-Block Diffusion (SingleBD) to Multi-Block Diffusion (MultiBD), where a running-set of consecutive blocks is decoded concurrently for inter-block parallelism. However, existing BD-LMs are mostly trained under teacher forcing, where the model observes only one noisy block conditioned on a clean prefix. While the recent di...

Speculative decoding accelerates inference by using a lightweight draft model to generate candidate tokens in parallel, and are then verified by the target model, enabling lossless acceleration. Recently, diffusion-based speculative decoding further improves parallelism by generating multiple tokens per forward pass via block-level diffusion, achieving state-of-the-art (SOTA) performance. However, existing methods adopt a fixed inference block size and assume a uniform optimal decoding strategy ...

Scientists have developed a non-invasive method to decode brain waves directly into words, eliminating the need for surgical implants in brain-computer interfaces. This breakthrough could revolutionize communication for individuals with disabilities.

The Department of Commerce has removed export restrictions on Claude Fable 5 and Mythos 5, allowing these AI models to be distributed internationally. This represents a significant shift in policy regarding advanced AI technology availability.

ArXiv is entering a new phase of development with enhanced features and improvements to its preprint sharing platform. The update aims to better serve the scientific research community in the digital age.

Claude Code's pricing has increased 5x, making it significantly more expensive for users who rely on the tool for development tasks. This major price hike may impact adoption and usage patterns among developers.

Google, the New York Jobs CEO Council and Urban Assembly hosted an AI summit for 150 education and industry leaders.