Cainew

DeepSeek has introduced vision capabilities to its AI models, enabling them to process and analyze images alongside text. This multimodal expansion allows DeepSeek models to perform tasks requiring visual understanding.

Learn how GPT-5.5 Instant improves ChatGPT’s health and wellness responses with stronger reasoning, better context, clearer communication, and physician-informed evaluations.

Open-source observability tool that uses AI agents to self-heal your software

A Claude Code plugin that makes Opus behave like Fable — completion, evidence, and verification enforced as procedure. Ships only what a Fable-vs-Opus comparison proved transferable.

Inference-native Tokenmaxxing Agent Harness for Loop Engineering

Open survey and evidence map for AI agent evolution, self-evolving agents, memory, skills, harnesses, benchmarks, and agent-swarm systems.

Finally, an inbox you'll look forward to. Agents sort your messages, draft your replies, and clear the grunt work behind the scenes, all in a client so well-crafted that email feels light, fast, fun.

Terminal kanban on plain markdown, with optional Today/Tomorrow planner, 24h agenda + calendar, and a live Claude Code agent view. Use only the panels you want.

An agent harness that compiles a model into one provably-correct, self-retargeting CUDA megakernel and self-tunes it past cuBLAS at batch-1 LLM decode, paper: https://arxiv.org/abs/2606.09682

Sales teams have been stuck with stale databases for 15 years. Jesse changes everything.. the first internet-wide search engine built for sales & marketing. Ask in plain English: "Find newly opened soccer facilities in the Midwest needing turf solutions." Jesse scans the live web and finds the right buyers in the market today. We are an anti database company, we don’t scrape and store stale databases and sell them at premium. Every lead is found fresh from the live internet and delivered.

Stop acting as the routing layer between AI and your actual work. Elvin is proactive AI that finds coordination work across your tools, handles the messy multi-step parts, and asks before taking action. It turns scattered context from messages, meetings, docs, and task tools into ready-to-approve drafts, follow-ups, updates, and next steps.

Claude Fable 5 Desktop 2026 – Ultimate AI Storytelling & Code Assistant Tool

An autonomous AI employee that lives in Microsoft Teams and does real work across 3,000+ tools: reports, reconciliations, approvals, recurring ops. Not a copilot that drafts and waits. It ships. Live today, $100 in credits, no card.

Turn any video into an interactive AI experience. With Agentic Videos, viewers don't just watch - they pause, ask questions, get real-time answers, and interact with the presenter inside the video itself. Viewers experience content in a fully personalized way. Creators gain a new world of insight into knowledge gaps and intent: a viewer who asks three questions tells you more than a thousand passive completions ever could. Now on the D-ID platform, built with industry-leading expressive avatars.

This project demonstrates a persistent agent memory layer built on Elasticsearch that achieves 0.89 recall, enabling AI agents to effectively store and retrieve past interactions. The implementation provides reliable memory management for stateful agent systems.

Juno is a local, open-source voice writing app for Mac. It is the only voice dictation tool with live transcriptions. Speak naturally in Mail, Slack, Notes, Cursor, or the app you’re already using; Juno writes clean text, rewrites selected passages, uses snippets, and creates Notes, Reminders, and Alarms. No login, runs offline and free forever.

Genie was built on the belief that to truly "get" you, a meaningful social product must also "get" the people in your life. introducing Genie Mentions: the first AI that treats your circle as part of who you are. your taste. your friends' taste. what you're all into. Genie keeps you updated on big moves your friends are making, trips they’re taking, dreams they’re conjuring. if you want AI for work, ChatGPT awaits you. if you want to know what's going on in your world, tag Genie in.

We show the standard basis of transformer hidden states already provides a training-free, architecture-general feature basis. Individual dimensions encode semantic content via their signs (+/-1) and confidence via their magnitudes, acting as independent binary registers; a feature is a subset of dimensions with a consistent sign pattern, read by counting sign agreements with no learned rotation. We validate this Bag of Dims framework across seven models spanning language (Qwen 3.5-4B, Gemma 3-4B...

Deploying multimodal foundation models as closed-loop policies increasingly requires conditioning actions on observations that are no longer visible. However, existing benchmarks either expose the full state, conflate hidden-state reconstruction with other agent skills, or test recall only after an episode has ended. We introduce RNG-Bench (Reconstructive Non-Markov Games), a benchmark suite designed to isolate a base model's ability to reconstruct past observations and act on them during multi-...

Turkish is agglutinative: meaning is carried by morphemes, yet the subword tokenizers that drive modern language models split words by corpus statistics, fragmenting semantically loaded suffixes and -- in the case of WordPiece and rule-based analyzers -- failing to decode their output back to the original text. This paper presents Morpheus, a neural morpheme-boundary model for Turkish that is at once a lossless, morphology-aware tokenizer and a word-embedding producer. A differentiable Poisson-b...

Score- and flow-matching models often rely on preference-based reinforcement learning for two purposes: aligning with subjective preferences and, surprisingly, recovering properties such as visual realism and coherent object structure that matching-based training is intended to learn from the data itself. We argue that this reflects a structural mismatch. Matching losses measure ell_2 regression error on the velocity or score field under training-time marginals, a proxy poorly aligned with the v...

Passive models for long video understanding typically rely on a "watch-it-all" paradigm, processing frames uniformly regardless of query difficulty, causing computational cost to grow with video duration. Although interactive frameworks have emerged, they often rely on global pre-scanning, and their context cost still scales with video length. We propose OmniAgent, the first native omni-modal agent that formulates video understanding as a POMDP-based iterative Observation-Thought-Action cycle. O...

On-policy self-distillation (OPSD) trains a model on its own rollouts and uses a frozen copy to provide dense token-level targets conditioned on a reference target. This works well for LLM reasoning, but a direct extension to multimodal large language models (MLLMs) can create a shortcut: the privileged target may guide tokens mainly based on the text reference target rather than the image. We propose ViGOS, a visually grounded OPSD framework for MLLM post-training. The student first writes a vi...

Diffusion models have become a promising alternative to autoregressive models. Among these, uniform diffusion language models (UDLMs) permit any token to be updated at any step, in principle enabling more flexible generation. However, no UDLM has yet been pretrained from scratch at both large parameter scale and large token budget. Both autoregressive modeling and masked diffusion modeling already have capable models at scale that the community can study and build on; uniform diffusion has none....

Learning to simulate human users in interactive settings could advance the training of agent assistants, evaluation of personalization systems, research in the social sciences, and more. Existing approaches generally do so by training a large language model (LLM) to match a single ground truth response, either by maximizing the log probability or by using a similarity reward. We instead propose {Turing-RL}: a Turing-Test-based reinforcement learning approach for training user simulator models. {...

Reinforcement Learning with Verifiable Rewards algorithms like GRPO have emerged as the dominant post-training paradigm for complex reasoning in LLMs, yet commonly suffer from policy entropy collapse during training. We conduct a first-order gradient analysis of token-level entropy dynamics under GRPO and identify a token-level credit assignment mismatch: the per-token entropy variation decomposes into the product of the trajectory-level advantage and an entropy sensitivity function over the nex...

Multi-turn tool-use RL is bottlenecked by the rapid depletion of informative samples in static datasets. We observe that the gradient signal in GRPO concentrates on tasks with the highest rollout reward variance, a consequence of the Popoviciu upper bound. Consequently, samples near the agent's capability boundary -- where successes and failures are roughly balanced -- contribute disproportionately large policy gradients. As training progresses, this boundary continuously shifts, which gradually...

Reinforcement learning (RL) has become a representative post-training paradigm for LLMs, enabling strong reasoning and agentic capabilities. However, rollout generation remains a dominant latency bottleneck because autoregressive sampling decodes responses sequentially and a small number of long-tailed generations often determine completion time. Speculative decoding (SD) offers a natural way to address this bottleneck, as it is a well-established technique for serving fixed LLMs that reduces la...

AI systems can increasingly automate scientific workflows, but the reasoning that links prior evidence, generated ideas, experiments and final claims often remains implicit inside model inference. Here we introduce Xcientist, a research harness that externalizes research synthesis and experimental validation into inspectable, contract-governed processes. Xcientist organizes literature evidence, idea states, implementation plans, ablation records and repair traces as persistent research artifacts...

Video generative models ( VGMs) have become a new frontier that can be used not just for video generation but for a multitude of downstream tasks, including world modeling. To advance these tasks, a good video model must understand the physical reality of the world. Evaluating this understanding is an emerging field and has led to the Physics-IQ benchmark, which quantifies this explicitly by comparing model-generated videos to real-world videos of physical experiments. In this work, we present a...

Motion forecasting is central to visual intelligence: agents must anticipate how objects will move in order to plan actions, reason about physical interactions, and synthesize realistic futures. We argue that 3D points in world coordinates provide a general representation that is class-agnostic, view-stable, compact, and directly useful for downstream tasks. We formalize the task of goal-conditioned 3D point motion forecasting: given a short visual history, a set of 3D query points on an object ...

Jun 18, 2026Frontier Red TeamProject Fetch: Phase two

This piece argues that local Qwen models should not be directly compared to Claude Opus as inferior, but rather recognized as serving different use cases and purposes. Local deployment options like Qwen offer distinct advantages for certain applications despite different capabilities.

Security researchers have demonstrated that ChatGPT's image generator can be manipulated through prompt engineering to produce violent and sexual content that violates OpenAI's usage policies. This vulnerability highlights potential risks in multimodal AI safety.