AI News

We're deepening our commitment to scientific research and technological discovery by committing $150 million to the Genesis Mission.

We'll also make Claude and our technical support available to more than 15 federal agencies.

https://t.co/MUmQBiC6MF

The financing includes an $8M seed round co-led by Sequoia Capital and First Round Capital, and a $30 million Series A led by Hummingbird Ventures.

Databench by Alkera is the open-source, multiplayer workspace for data science, analytics, and engineering. Collaborate live alongside teammates and agents in notebooks and chats, run any cell or agent on your laptop, another computer, or a GPU node and launch many agents in parallel to explore ideas. Every result traces back to the data and code behind it. Try it at https://alkera.ai with a generous free tier or host Databench yourself (https://github.com/AlkeraAI/Databench).

Manus is a general AI agent that turns your thoughts into actions. It excels at various tasks in work and life, getting everything done while you rest.

Teradyne Robotics and Elite Robots have settled their software dispute, ending the legal case without disclosing the terms of the agreement.

NVIDIA RTX Spark PCs are now available for pre-orders, bringing the best of NVIDIA's AI and gaming capabilities within sleek and slim designs.

Long-horizon reasoning demands access to earlier information at a manageable generation cost. Full-history attention incurs growing storage and computation, while recurrent compression can lose precise details. Therefore, we present YANchor-4B, a general-purpose recurrent model that preserves crucial memory as ANchors for retrieval during subsequent reasoning. Beyond $O(N)$-time generation and $O(1)$ memory, YANchor enables effective long-horizon reasoning through its multidimensional memory mechanism. For example, on challenging math problems, it achieves 82.93% mean pass@1 on AIME 2024--2026 and 63.64% on HMMT, substantially outperforming linear-time, constant-state counterparts, including larger models. It also delivers several-fold higher batched long-generation throughput than Transformer and hybrid baselines on H100. Furthermore, evaluations across dozens of benchmarks demonstrate YANchor's superiority in general-purpose capabilities.

Join the discussion on this paper page

Most voice assistants work like a relay race: speech becomes text, text goes to a model, the answer becomes speech again. Every handoff adds a pause. ⏱️

Speech-to-speech models skip the relay. Google's new Gemini 3.8 Live models listen, reason, and respond in one system.

🎯 The Extended Thinking …

SynthID Detector is now available to everyone. 🌐

Check whether online content was generated using @GoogleAI, or with tools from our industry partners – including @OpenAI, @NVIDIA, Kakao and coming soon, @Apple.

Try it out → https://t.co/cPW2aNnqr6

OpenBot is an open-source, local-first runtime that lets you drop diverse agents into a single workspace and orchestrate them using a clean tag-and-channel system. Just tag their capabilities and let OpenBot dynamically route tasks, sync collaborative workflows, and manage persistent threads right in your local file system. Fully modular, completely self-contained, and built for developers and makers who want absolute structural control over their agent fleets.

Cloudflare is a leading edge network services provider that offers a wide range of solutions to enhance the security, performance, and reliability of websites and applications. With its global network infrastructure and advanced technologies, Cloudflare empowers businesses to build a faster, more secure, and resilient online presence.

Sigil Wen, backed by a Silicon Valley who's who, has built an on-device AI assistant that promises to be free, fully private, and capable for everyday tasks.

A new personal AI assistant has arrived. Meet Tab, which announced Tuesday that it had officially emerged from stealth.

Mecka AI collects and analyzes human motion data to train humanoid robots and other kinds of robots. The startup pays people to record everyday tasks.

Your marketing campaigns are personalized. Your landing pages should be too.

GenPage learns your brand, builds dedicated pages for every ad, keyword, or target account, then continuously A-Z tests them on Autopilot - turning more traffic into pipeline and revenue.

IrisGo is an AI operating system for Windows and Mac that helps automate everyday work across applications. With Watch & Learn, demonstrate a task, turn it into a reusable workflow, and run it again with new inputs. Get daily briefs, surface important emails, and keep your next actions in view. Spend less time on repetitive work and more time on what matters.

Faraday Future is planning to bring a robotics factory online in the United States by the end of 2026 as the electric vehicle company accelerates its expansion into humanoid and other embodied AI r…

We study whether latent sequence compression of prompt prefixes preserves the capabilities that large language models rely on during inference. We introduce SemanticFold, a compression scheme that folds prefix hidden states at learned boundaries, and evaluate it across five model scales: Qwen3-1.7B, Qwen3-8B, SmolLM2-1.7B, Pythia-1.4B, and Pythia-6.9B. We use a fixed-target protocol: a frozen prefix is executed natively or compressed, and both arms teacher-force identical continuation tokens. This design rules out target-selection explanations for likelihood changes. We examine five endpoint families: fixed-target negative log-likelihood, finite-label reasoning accuracy, linear probe accessibility, open-ended generation, and systems-level memory and latency. We find that compression moves these endpoints non-monotonically and that they do not share a single compression threshold. On Qwen3-1.7B at compression ratio R=1.7, compressed-minus-native mean NLL decreases by 0.135 under paired bootstrap with 10000 draws. On SmolLM2 at R=1.2, the mean change is 0.013 higher than native. On both Pythia checkpoints, NLL is effectively unchanged. An NLL decomposition separating sequence shortening from the learned residual transform shows that the favorable Qwen likelihood is attributable primarily to residual adaptation rather than to shortening alone. MLP-only, which applies the transform without shortening, achieves 0.082 lower NLL than Full SemanticFold. Linear probe accuracy and macro AUC change by less than 0.03 in absolute value across conditions, with confidence intervals crossing zero. We conclude that preservation under latent compression has no single scalar certificate: language-model fit, decodability, and reasoning behavior answer different questions and can move in different directions under the same compression operation.

Conditional memory architectures such as DeepSeek Engram use input n-grams to look up learned embeddings, expanding the capacity of large language models (LLMs) with limited additional computation. Beyond model scaling, this architecture has demonstrated the potential to decouple factual knowledge storage from general-purpose computation, offering a promising route to updating factual knowledge while keeping the Transformer backbone fixed. Realizing this potential is challenging because different expressions of a fact may activate different n-gram embeddings, while updating shared embeddings can unintentionally change the model's predictions about other facts. We propose EngramEdit for decoupled knowledge updates through conditional memory. EngramEdit first computes target memory representations that make the model predict the updated fact across multiple expressions. It then jointly updates the shared n-gram embeddings to match these targets across expressions and edits, penalizing updates to frequently reused embeddings more strongly to preserve unrelated knowledge. Experiments show that EngramEdit enables independent factual knowledge updates through conditional memory, achieving near-perfect editing success. Revised knowledge is usable across unseen expressions and in multi-hop reasoning, with nearly three times the strongest baseline's accuracy under chain-of-thought (CoT) prompting. Unrelated knowledge and general capabilities are largely preserved even as factual updates accumulate. These findings show that EngramEdit turns conditional memory into an editable knowledge interface, extending its role beyond model scaling to support decoupled knowledge updates.

Terminal-agent capability depends jointly on model weights and the runtime harness that formats prompts, binds tools, and handles error recovery. Existing harness-model co-evolution approaches improve both components, yet often treat trajectories produced during harness search as an undifferentiated replay buffer. This practice overlooks that a trajectory's value for model training depends on the harness under which it was generated. To systematically analyze this interface, we establish an alternating co-evolution framework that decouples harness search and policy training through component-wise promotion decisions. Within this framework, we introduce CoTrace, a harness-aware data recipe that explicitly governs trajectory routing, provenance matching, and curriculum refresh. Under CoTrace, recurring execution failures guide harness synthesis, while policy training is strictly conditioned on verified rollouts matched to the adopted runtime for supervised fine-tuning (SFT) or fresh online interactions for reinforcement learning (RL). On the Tmax promotion split, CoTrace advances Qwen3.5-9B from 78 to 88 solved tasks under supervised fine-tuning while an online reinforcement variant reaches 90. Specifically, a compact harness-matched corpus produces steady model gains at substantially lower compute than much larger corpora pooled across sibling harnesses. Furthermore, evaluations on Terminal-Bench 2.1 and SWE-bench Lite show that out-of-distribution transfer depends fundamentally on harness compatibility, where maintaining consistency between training and evaluation runtimes prevents procedural execution breakdowns observed under foreign scaffolds.

SpaceX is racing to raise tens of billions in borrowed money to lock in an exclusive chip supply deal, and the timing raises questions about who really benefits when the money moves.

GPT-6 and Intelligent UI, now rolling out in ChatGPT for everyone.

Intelligent UI in ChatGPT delivers fast, interactive answers that make everyday questions more visual, complex topics easier to grasp, and interactive tools for your task available on the spot.

devpit is a native desktop app for running Claude Code agents. One terminal you actually watch, a board where moving a card runs real work, a cost on every agent call, and an orchestrator that sees every project at once. Local-first, open source.

FastRouter is a unified AI gateway and control plane for developers and enterprise teams building with LLMs. It routes every request to the right model across 200+ LLMs through one OpenAI-compatible API, optimizing for cost, latency, quality, and reliability. With intelligent routing, failover, observability, and governance, teams can scale AI apps without vendor lock-in or code changes.