OpenRouter BlogOpenRouter tested three methods for obtaining structured outputs from LLMs, validated on nine models supporting structured outputs and only tool-supported GLM 5.3 FlashX, where 209/216 responses matched strict JSON Schema.
OpenRouter BlogOpen-source multimodal model Claude Code v2.1.159 released, supporting multilingual and code generation.
NVIDIA Technical Blog6565The NVIDIA Alpamayo 2 Edge edge inference model can run within automotive computing and latency constraints. The model achieves 70.5 on the MMLU benchmark, supports an 8k context window, speeds up 3.5x, and costs $3/ million tokens with closed-source deployment.
LangChain BlogLangChain improves AI agents through evaluation-driven development (EDD), defining agent behavior standards, building evaluations, adjusting code, and comparing results. EDD draws inspiration from test-driven development (TDD), targeting agents with non-deterministic outputs, measuring performance through evaluation metrics, and guiding development. Specific steps include defining agent behavior specifications, building test datasets containing inputs and reference outputs, and writing code or LLM evaluators to verify whether the behavior meets expectations.
OpenAI News5050OpenAI Academy will provide free ChatGPT practical workshops for caregivers in five major cities in the United States, helping them master practical AI skills for daily life.
A user developed a browser-based tool that allows inputting a location to query all satellites passing over it in the past 7 days and their real images (Sentinel-2/Landsat/NASA), rendering approximately 17,000 on-orbit objects via GPU and calculating Kepler's equations in real-time to achieve high-precision positioning. The tool only displays real satellite data and does not store user locations.
Browser Use1Password connects to Browser Use via a service account, enabling automatic password filling and TOTP code completion for web tasks. Create a dedicated password vault and configure read-only permissions for the service account. After connecting with a Browser Use token, running login tasks can verify the authentication page. The integration requires an active 1Password connection for each project, distinguishing between service account tokens, vault IDs, and Browser Use API keys.
Meta Engineering3838LLDB supports application debugging across CPU and GPU through the neutral accelerator plugin architecture of lldb-server, separating general debugging infrastructure from hardware-specific device control, presenting CPU and GPU as independent debugging processes that work together in a single debugging session. This architecture converts hardware-specific debugging APIs into concepts understood by LLDB through plugins, supporting collaborative debugging of CPUs and GPUs with different architectures, registers, memory spaces, and execution models while maintaining the neutrality of the client.
Engineer IainKay developed the open-source script Ram-Altman, which monitors computer memory usage through the Linux system. When available memory exceeds 10% of the total capacity, it automatically sends an email to OpenAI CEO Sam Altman daily to report idle RAM amounts, and has prepared various email subjects, body texts, and signatures. The project has been uploaded to GitHub for free use, but the developer warns users to avoid mass emailing to bother Altman.
Developer brunostjohn successfully ported iPadOS 27.2 natively to the Nintendo Switch and enabled Liquid Glass effects, running directly on the Switch's NVIDIA Tegra X1 chip, with some scenes reaching 60 FPS, but the GPU driver still needs fixing, hardware compatibility is about 90% complete.
When working with Claude Code or Codex from the terminal, users often experience mental exhaustion due to multi-threaded task processing. The author manages tasks in git by organizing work in a way similar to a priority queue, with each task being a markdown file and its status as a git snapshot. The AI agent processes new tasks in a loop, runs tests, and performs browser checks, but cannot make decisions and requires the user to handle them in the 'Owner Inbox' by viewing tasks with the Claude tool on a single page, closing them after completion to process one task at a time.
Simon Willison6565Claude Code or Codex generated PR descriptions are too verbose, but hiding them in HTML details/summary elements can optimize the reading experience. Use the model name and effort level as the summary, for example, 'Claude Opus 5.5 Medium', so readers can quickly get the core information. GitHub Markdown supports this feature; click the expand arrow to view the full content, making the PR no longer appear messy.
Latent Space4242Stanford University's CS146S course has changed assignments to contributing to open-source projects, letting students use AI code review tools and learn collaboration; Monaco company adopts Devin code review tool, while building its own Agent workspace to maintain technical autonomy, and develops Hal system, allowing engineers and salespeople to solve problems through Agents.
llama-modchip releases a set of ggml/llama.cpp core patches and quantization transcoding tools, compressing the 284.33B DeepSeek-V4-Flash to 56.56 GiB (1.71 bpw), which can run on a pure CPU machine with 64 GB memory, with a peak memory of 57.61 GiB under 256K context and no need for swap.
Why it matters: The project provides a quantization scheme and core patches for running 284B MoE on 64 GB memory, along with measured data for each hardware, which can be used to determine the feasible boundary for pure CPU deployment.
Qualcomm showcased its edge-side MoE intelligent agents at the 9th CIIE, driving edge-side AI from single-point applications to intelligent agent experiences. MoE technology enhances the intelligence level of edge-side models through mixed expert modules, achieving higher complexity task completion in the 2 billion to 3 billion parameter scale. Qualcomm collaborated with Infris AI, Changelong, and Jieyu Xingchen to optimize models, inference, storage, and chip platforms, demonstrating a systematic solution for edge-side AI ecosystem synergy.
After the Nintendo 'The Legend of Zelda: Ocarina of Time' remastered version launched on Switch 2 on November 5, developers have successfully run it on jailbroken PS5 and iPhone. The PS5 version is based on the EglantinePS5 project, with a frame rate of about 15-20 FPS under the 12.40 firmware, but there are memory issues and rendering problems, with the patch written by Claude.
MarkTechPost6565Use llama.cpp to run the NVIDIA Dory 1.0 4B model, which achieves inference enhancement through repeated intermediate layer calculations. Each additional loop increases computational load without increasing parameters. Build the specified branch of llama.cpp, download the 3.7 GB GGUF file, analyze the parameters and cache consumption per loop, test decoding speed and cache impact, and evaluate accuracy, answer length, and inference time changes with loop counts on GSM8K.
The Verge · AI6565The NYT Cooking AI Assistant provides practical substitution suggestions and cooking guidance by analyzing user reviews, helping users adjust recipes based on available ingredients. The assistant was launched for the iOS version, where users can input questions in a small text box on the recipe page, such as 'What can I substitute for this ingredient?' or 'Adjust the recipe for seven people, making it vegetarian and gluten-free,' AI will generate a new version of the recipe and highlight the changes. Users can switch back to the original recipe at any time; the assistant does not modify the original recipe but creates a new version.
GitHub Blog · AI & ML3838GitHub Copilot supports cross-device session continuity on canvases, allowing users to execute tasks in cloud sandboxes, isolate repository editing through the Copilot app, enhance efficiency with storage instructions and shared memory project knowledge, and optionally match work requirements with the Copilot Fusion model.
Claude · YouTube4242Salesforce expanded Claude Code to thousands of engineers, doubling effective output and reducing the number of bugs per line of code. The 'effective output' scoring system, co-built by the company's engineering leadership and Stanford researchers, measures AI coding ROI, and the intelligent agent maturity curve guides engineers from simple prompts to fully autonomous agents.
O'Reilly Radar5858Claude Code provides detailed control over the code implementation process through CLAUDE.md instructions, skills, sub-agents, and hooks. Hooks can interpret reject operations or return compilation or policy failures, but external lifecycle control still requires other layers.
Browser UseClaude Code v2.1.159 released, supports 8k context window, performs excellently on SWE-bench Verified; Anthropic launches Browser Use open-source library, allowing AI agents to control browsers via Chrome DevTools Protocol, providing Browser Use Cloud and Browser Use Agents services.
O'Reilly RadarAt the recent Foo Camp, participants believed that the primary carrier of software development is shifting from executable code to natural language prompts, generative processes, and developer skills. Some compared Python to assembly language, suggesting that natural language generation might make traditional code writing optional.
NVIDIA Technical Blog5555NVIDIA DSX OS can configure, authenticate, and operate Kubernetes GPU clusters, addressing the issue of slow distributed training tasks due to poor performance of individual GPUs. The system ensures cluster stability by monitoring performance at various stages of the cluster lifecycle, such as node replacement and driver patching.
LangChain BlogLangChain explores three principles of human feedback to enhance agent development: expert judgment defines quality standards, comprehensive context evaluation, and converting feedback into reusable data; through agent trace to record user interactions and model behavior, manually labeled scoring with reasons, used to calibrate automated evaluators, improve evaluation criteria, and build golden datasets, forming reusable expert knowledge assets.
IBM Research Blog3838HRL Laboratory develops an inertial measurement unit (IMU) using advanced algorithms, providing an uninterrupted GPS alternative. Its MEMS inertial sensor is the size of a can of soup, with accuracy close to GPS, and solves the long-standing problem of bias drift. Through the AutoCal automatic calibration system, it continuously corrects errors, achieving unlimited navigation support. The IMU uses ASIC and advanced packaging technology to optimize cost and size, making it suitable for light vehicles such as drones.
Netizens Successfully Bypass iFIT Subscription Wall on NordicTrack Running Machine Using Claude Code, Transforming the Device's Large Screen into a Free-Use Streaming Interface. The Modification Process Took About 1.5 Days to Complete, and the Project Has Been Opened on GitHub, Integrating Apps Like Netflix, Crunchyroll. Netflix Launch Time Is Reduced to About 21 Seconds, but Only Supports Standard Definition Quality and May Affect Warranty.
The TVPort open-source project, based on the nxnative toolkit and Claude Opus 5.5, allows the Apple TV 4K to natively run Switch game 'Super Mario: Odyssey', achieving 60FPS performance with occasional minor stuttering. The project requires users to extract game data themselves and use modified Switches, not a solution for average users.
NVIDIA Technical Blog6565NVIDIA has developed the AI Factory Operations Agent to simplify operations within DGX SuperPOD, automatically correlating compute, network, hardware, and scheduling issues through a single interface to help SREs identify the causes of training performance degradation. The agent can process data from multiple dashboards, improving operational efficiency.
Claude · YouTubeCreate and share documents, trackers, and dashboards in Claude, allowing teams to collaborate on edits from the chat interface; set access permissions from viewers to editors, and distinguish between commenter and contributor roles; supports automatically executing updates to comment content or styles via Claude.
AWS Machine Learning BlogFeatured6060The AWS team, referencing the open-source multi-agent data analysis solution Deep Insight, has broken down the context engineering of production-level AI agents into four layers: multi-agent architecture isolation of context, using prompts to limit output, externalizing heavy content to the file system, and using a validation agent with a fallback mechanism.
Why it matters: The AWS team has broken down the context engineering into four layers and provided specific code and token-saving data from the open-source reference implementation, which can be directly used to modify your own Agent.
NVIDIA Technical Blog6565NVIDIA NeMo can transform agent failures in production into priority issues and generate customized evaluations to help improve prompts, tools, models, memory, skills, and code. When building agents for specific domains, general benchmark scores may not explain their failure reasons, such as the agent choosing the wrong tool, following the wrong workflow branch, omitting relevant context, or using unnecessary high-cost models.
O'Reilly Radar6565Does AI make software development faster? Delivery speed is not productivity, but the ratio of value output to cost. Since software deliverables are not standardized, metrics like lines of code and PR numbers are unreliable, and estimations are inaccurate. Measuring delivery speed requires comparing a team's performance before and after. Through randomized controlled trials, alternating between tasks using AI and tasks not using AI can reduce interference from other process improvements, thus more accurately assessing the actual impact of AI on delivery speed.
Cohere BlogCohere's Embed 5 and Rerank 5 have been launched on OpenRouter, allowing workflows to be built using a single API through just a few lines of Python code. Embed 5 Pro and Fast share embedding space, with Pro for indexing quality and Fast for query latency, while Rerank 5 scores are calibrated across queries and domains.
Open-source project Megler runs Android apps on Windows 10 Mobile 15063 and higher versions without a subsystem.
