Harvey's agents now inherit the context of your matters and projects, along with a memory of how you work.
Introducing Harvey Tenet, our first post-trained open-weight model.
550 points, 852 comments
We ran 904 DeepSWE rollouts on GLM-5.3 and GPT-5.6 Sol. Sol leads pass@1 by 3.7 points; GLM-5.3 wins pass@4 at half the cost, and a GLM-first cascade hits 85.9%.
We ran 904 DeepSWE rollouts on GLM-5.3 and Claude Fable 5. A tie on pass@1, but GLM-5.3 wins pass@4 and costs 5.4x less: \$3.99 per rollout vs. \$21.63.
250 points, 196 comments
Amazon Bedrock now offers OpenAI GPT-5.6 models (Sol, Terra, and Luna) in more than 25 AWS Regions with cross-Region inference. Learn how US geographic and global inference profiles route requests for higher throughput, how to call the models with the OpenAI and Converse APIs, and how to configure IAM, quotas, and monitoring.
Healthcare, retail, and life sciences teams store large volumes of operational data in Snowflake, but turning it into predictions is hard. In Part 1 of this series, you set up your AWS account and Snowflake environment for a no-code ML workflow with Amazon SageMaker Canvas, laying the foundation for building a fraud detection model without writing code.
In Part 2 of this no-code ML series, you connect Amazon SageMaker Canvas to Snowflake, prepare and join transaction data with Data Wrangler visual transformations, and train an XGBoost fraud detection model. All without writing machine learning code, laying the groundwork for interactive dashboards in Part 3.
In Part 3 of this no-code ML series, you bring fraud detection predictions to life. Import your Amazon SageMaker Canvas predictions into Amazon Quick Sight, build interactive dashboards, use generative BI to answer questions in natural language, and publish AI-generated executive summaries for stakeholders.
Deploying vision-language models (VLMs) on mobile devices is challenging due to their significant memory and compute requirements. We present a framework for quantizing VLMs for efficient inference on resource-constrained hardware. Our approach combines a quantization pipeline that uses the model itself to generate training data and does not require access to the training setup, with a novel 2.7-b
Agents learn to act through interaction with environments, yet the environments used for training are often manually constructed or synthesized around predefined tasks and benchmarks. This task-centric paradigm makes it difficult to scale environments that reflect realistic and evolving workflows where diverse tasks can naturally emerge from the underlying world. We introduce AgentMercury, a scala
Simulating realistic user shopping behavior underpins offline evaluation and reinforcement learning in e-commerce scenarios. While recent LLM- and VLM-based simulators have made encouraging progress, reproducing a real browsing session remains difficult for two reasons. (i) Memory Challenge: a shopping session spans dozens of pages, yet existing agents either discard long-range observation histori
Real-world image search queries are multimodal and compositional: ``find this shirt in pink'' specifies an entity to retain, an attribute to modify, and context to ignore. Yet existing re-rankers either compress such multifaceted relevance into an opaque embedding or rely on free-form chain-of-thought that easily omits or hallucinates fine-grained constraints. Drawing on rubric- and checklist-base
Reliable spatial understanding is an important prerequisite for future medical vision-language systems that aim to support radiological report generation and structured image understanding. While modern vision-language models (VLMs) show promising performance on many medical imaging tasks, recent evidence suggests they remain weak in controlled spatial reasoning and often fail to reliably ground s
Improving the safety of large language models (LLMs) often comes at the expense of utility, as globally applied safety tuning may affect model responses to both harmful and benign inputs. We propose Continuous LatEnt Adapter Routing (CLEAR), a conditional safety adaptation framework that uses a lightweight hidden-state gate to continuously control the activation strength of a safety low-rank adapt
We present PhysCaP, a Physics-Informed Code-as-Policy agent for active perception in robotic manipulation. While vision-language-action policies excel at imitating demonstrations, they rely on passive observation and fail to infer latent physical properties critical for manipulation. PhysCaP augments code-as-policy frameworks with a physics-informed exploration layer that enables explicit informat
Deep face recognition (FR) models reach near-saturated accuracy but remain opaque: a practitioner cannot ask which semantic attributes a similarity score relied upon. EXPL-FR answers this inside the FR model's own embedding space. A lightweight adapter aligns a vision-language model's (VLM) image encoder with the frozen FR space, trained on face images alone and never on text. Because the VLM's en
LLMs have evolved from language generators to autonomous agents capable of complex, long-horizon tasks. This evolution has produced paradigms including Prompt Engineering to elicit model capabilities, Context Engineering to manage information access, Harness Engineering to organize external tools and resources, and Loop Engineering to support continual reflection and self-improvement. Yet as tasks
Recent omni-modal large language models (Omni-LLMs) show great potential as real-time video assistants, which continuously perceive environments and guide users to achieve specific goals. Unlike traditional passive video understanding, interactive assistants should actively combine visual states, user goals, and prior knowledge to provide effective help. Evaluating this is rather challenging, as t
We define an atomic generation fact f=(u,tau,omega,z;rho), recording the origin, realized transformation, concrete occurrence, generated result and relation role. Compiled into a Generation-Fact Graph (GFG), these facts provide an AI-native, compilable scientific fact substrate preserving generation histories. We establish a GFG-based recursive scientific process in which analysis, intervention, r
Prompt injection is listed as the \#1 threat to AI agents. When an agent accesses external data from websites, files, or emails, an attacker may inject a prompt into the data, saying, "Ignore all prior instructions and perform ." To prevent arbitrary manipulation of agents, defenders try to train secure LLMs, which, however, still suffer from near 100% attack success rates (ASRs) against adaptive
E-commerce live streaming requires omni-modal understanding of noisy, temporally extended streams, where product facts are distributed across speech, video frames, product images, overlaid text, and user queries. We present TLive-Omni, an omni-modal understanding model tailored to live-commerce scenarios. It maps image, video, audio, and text inputs into a unified representation space. For long-fo
Serving large language models cheaply increasingly means shipping models that are both structurally compressed to a fraction of their parameters and quantized to 4 bits. Together these steps degrade reasoning, mathematics, coding, and long-context behavior enough to require a recovery, or healing, stage before deployment. The default recipe, quantization-aware training (QAT), re-fits the compresse
Repository-level coding benchmarks have become the standard for evaluating coding agents, yet they inherently suffer from data leakage because they are built upon popular open-source repositories repeatedly used for training. Consequently, strong performance may reflect memorization of canonical repository cues rather than robust repository reasoning. We propose SchrodingerRepo (Schrödinger's Repo
With large pretrained models, existing methods have effectively improved instruction-based video editing. However, most of them rely on an in-place editing assumption. They align the edited video with the given source clip frame by frame over a fixed time span. This pattern fails for open-ended streams, e.g., restyling a live game or applying a camera move to an ongoing shot. In such cases, edits
261 points, 145 comments
New MIT research could lead to better materials for a fossil-fuel-free process for making the chemical that's essential to fertilizer and other products.
Since announcing Google Antigravity in Gemini Enterprise Agent Platform at I/O in May, we’ve heard helpful feedback from our customers. Your developers want easy access to coding agents across surfaces. Your enterprise governance team wants security controls and license management. And your finance team wants pooled u
AI agents can take actions that do not match your organization's policies. Policy in Amazon Bedrock AgentCore lets teams enforce controls across agents, now including time-based constraints. This post shows how Policy Authoring turns natural-language policy documents into correct Dogwood policies, with worked examples and best practices.
Scaling agentic AI across an enterprise requires patterns that preserve flexibility while avoiding vendor lock-in. In this second post of our multi-agent series, we examine how ML teams operate many agentic AI systems across a multi-everything environment of frameworks, models, and providers, and the principles that let those systems scale together.
Learn how AWS Professional Services uses a multi-agent framework built on Amazon Bedrock AgentCore to automate enterprise cloud migrations end to end. Purpose-built AI agents handle discovery, infrastructure as code generation, portfolio governance, and post-migration operations, reducing IaC development time from weeks to minutes.
Software teams don't ship without linting, code review, and canary deploys. Release governance brings that same discipline to Sierra: the checks, approvals, and staged rollouts that move a change safely from a builder's Workspace to a live customer conversation.
AWS offers a broad portfolio of vector search built directly into the databases and storage services you already use, with no standalone vector database or data migration required. This post covers six purpose-built services, a decision framework for choosing the right engine, and customer proof points for each.
It’s never been easier to start an AI-powered startup on Google Cloud. You grab an API key from Google AI Studio
To satisfy the demands of enterprise-grade agentic AI applications, underlying vector databases often struggle to scale effectively as modern use cases can scale to billions of vectors. As a fully managed PostgreSQL-compatible database service,
Learn how to add context-aware security monitoring to FHIR APIs using Amazon Bedrock. This post shows how to detect anomalous access patterns, classify data sensitivity automatically, and generate compliance reports in natural language, all without adding latency to clinical workflows.
We present an oscillator-driven femtosecond transient microspectroscopy system based on supercontinuum generation in a photonic crystal fiber (PCF). The system operates at 80 MHz without pulse amplification, enabling high-sensitivity measurements with a detection sensitivity better than $10^{-4}$ while using low pulse energies suitable for microspectroscopy. Optimization of the PCF length ($\leq 5
Mixture-of-Experts (MoE) architectures significantly expand model capacity without a proportional increase in computational cost. However, optimizing their hyperparameters---particularly the learning rate---at extreme scales of both model size and token budget via sweeping remains computationally prohibitive. In this paper, we propose a compute-efficient, two-step hyperparameter transfer framework
967 points, 526 comments
Long-context modeling is a pivotal capability for Large Language Models, yet the quadratic complexity of attention remains a critical bottleneck, particularly during the compute-intensive prefilling phase. Our previous work, FlashPrefill, mitigates this cost through instantaneous pattern discovery and max-based dynamic thresholding; however, it remains an algorithmic prototype that is still distan
With a fixed deadline and design resources committed elsewhere, Stampli used Codex and ChatGPT Work to compress weeks of launch production into days.
Introducing Harvey Tenet, our first post-trained open-weight model.
We characterize the learning dynamics of a compact hybrid quantum forecasting model through comparison with a structurally aligned classical baseline. Using stationary harmonic-mixture and nonstationary chirp benchmarks with controlled spectral complexity and data availability, we analyze empirical Neural Tangent Kernel dynamics through kernel-target alignment, kernel drift, spectral concentration
We present a catalog of 147 short-period (10.34--106.46~min) blue compact-binary candidates, identified by combining Gaia DR3 astrometry and photometry with ZTF DR23 light curves via a Gaia selection, period searches, and machine-learning morphology ranking. Of these, 111 lack prior compact-binary classifications. Multiwavelength data (DESI DR1, GALEX, AllWISE) reveal a heterogeneous sample: on th
Web Search on Amazon Bedrock AgentCore now supports runtime domain and published-date filtering. New per-request filters give developers per-call control over which web sources their agents consult and how fresh those sources must be, all enforced server-side. This release also expands Web Search to the Europe (Ireland) and Asia Pacific (Tokyo) Regions.
Classifying, extracting, and validating high volumes of documents is a challenge across banking, insurance, healthcare, and the public sector. See how a mid-size mortgage lender automates its entire document intake pipeline, from email to validated data, using the AWS GAIIC IDP Accelerator and Amazon Quick Automate.
In this post, you learn three serverless patterns (task-token callback, direct service integration, and durable functions) for invoking Amazon Bedrock AgentCore agents asynchronously from AWS Step Functions pipelines, eliminating idle compute costs while your AI agent processes requests.
9551 articles sourced historically · 100 per page