#部署/工程

Try our new AI that was built with it's user in mind

Olo is a Mac intelligent companion, free with in-app purchases, that answers questions from sources you can open and checks each sentence as supported, contradicted or not covered. It can be invoked with Option+Space in any app or via Services > Ask Olo on selected text, and it accepts PDFs, Word documents, notes and chosen folders. Olo Starter (£19.99) and Olo Growth (£49.99) are auto-renewable monthly subscriptions, and the app requires macOS 14.0 or later.

Unity Catalog Metric Views, now in Power BI: one unified, governed semantic layer for all your reporting and AI

Databricks introduces Publish to Power BI, which connects Unity Catalog metric views to Power BI semantic models so each metric view becomes a table, its dimensions become columns and its measures become DAX measures, with Power BI querying the metric view itself. Publishing can be done on demand from Catalog Explorer or automated via Power BI tasks in Lakeflow Jobs, and DirectQuery queries honor Unity Catalog access controls. The feature is available in Private Preview.

How Baseten and Genspark trained a deep research model to outperform Opus 5.5

Genspark worked with Baseten Applied Research to post-train Qwen3.5-397B into a specialized Deep Research model that scores 0.759 on Genspark's quality grader, up from 0.612 for the base model and above Opus 5.5 (0.738), Sonnet 5.5 (0.742), GPT-5.6 Sol (0.637), Kimi K3 (0.724) and DeepSeek V4 Pro (0.675). The recipe splits the task into research, trained with reinforcement learning against a shaped semi-verifiable reward, and synthesis, trained mainly with supervised fine-tuning on high-quality reports written from the model's own research trajectories, followed by end-to-end RL. The model crawls 5× more sources per query than the base model and runs with NVFP4 weights on 2×B200 GPUs per replica, now serving as the default for Genspark's Deep Research agent in production.

Autonomous treasury agents: How Trovata built an AI-native solution on Amazon Bedrock and AgentCore

Trovata, an AI-native cash and liquidity management platform connected to over 10,000 banks, describes how it built autonomous treasury agents on Amazon Bedrock and Amazon Bedrock AgentCore, using role-based access control and a verified API layer within a bank-branded deployment model. The post outlines use cases such as cash forecasting, natural language querying, anomaly detection, automated reconciliation and persistent agentic workflows, and cites customer results including $1M+ in annualized cost savings, 5+ hours per day saved per accounting team, a 77 percent reduction in annual TMS costs, and reporting reduced from hours or days to under a minute. It also notes a library of ten agent patterns from real customer deployments and Trovata's plans to expand agent patterns, ERP integrations and global bank connectivity.

AI and the Art of Work

A developer describes feeling drained by AI discourse and by agent-driven work that fragments the day into parallel async tasks, eroding the deep coding sessions he once valued. He explains how a daily pen-and-paper notebook spread, including a page tracking AI conversations and reviewing whether they helped, helps him stay intentional and grounded. He ends by inviting other humans to share what works for them, rather than more benchmarking noise.

Show HN: Relay – A multi-agent pipeline for AI-assisted front end engineering

Relay is a six-role multi-agent pipeline (PM, Architect, Implementer, Reviewer, Prod-Readiness, Verifier) for shipping frontend changes with AI assistance, with a human approval gate after the Architect and a Verifier that defaults to FAIL. It is tool-neutral via role contracts in .agents/, claims >90% cache hit on warm sessions, and offers a lighter /pipeline-lite mode that skips PM, Architect and Verifier but refuses tasks mentioning auth, payments, PII, tokens or cookies. It ships as a Claude Code plugin with slash commands, a GitHub Copilot paste workflow, and an AGENTS.md route, tuned for React 18, TypeScript 5, Vite, CSS Modules, Axios, Jest + RTL and Playwright.

Introducing Scribe v3 and v3 Live, our most accurate speech to text models

ElevenLabs launched Scribe v3 for recorded audio, which makes 27% fewer word errors than Scribe v2 across 102 languages and has the lowest diarization error of any model tested, and Scribe v3 Live for realtime audio, which makes 43% fewer semantic word errors than Scribe v2 Realtime and adds speaker diarization and improved turn detection. Scribe v3 Live is available today in ElevenAgents, and both models are available in ElevenAPI, where switching from Scribe v2 is a one-line change to the model ID (scribe_v3 or scribe_v3_live). Both run in the US by default with isolated EU, India, and Singapore environments, Zero Retention Mode, and SOC 2, ISO 27001, PCI DSS Level 1, and HIPAA-eligible coverage.

Pay for Hugging Face Hub with your Google Cloud Account

Hugging Face organizations can now link to a Google Cloud account through the Google Cloud Marketplace so that Hugging Face usage appears on the existing Google Cloud invoice instead of a credit card. The setup involves subscribing to the Hugging Face Hub listing, selecting the organization to link, and waiting for Hugging Face team approval, after which the Google Cloud account becomes the active payment method. The article notes that included Team or Enterprise credits still apply before pay-as-you-go usage is reported, and that linking and unlinking are recorded in audit logs.

Amazon Bedrock tokenomics data lake

AWS describes a centralized observability data lake that pulls Amazon Bedrock model invocation logs from linked accounts into a central account via CloudWatch subscription filters, processes them through Lambda and Amazon Data Firehose into Parquet in S3, and exposes them through Athena as a single queryable table. Per invocation, the flattened table captures the prompt, response, input and output token counts, model, stop reason, account, timestamp, and IAM role, and can be joined to CUR 2.0 data on the IAM ARN for per-invocation cost attribution. The solution is deployed with two CloudFormation templates from the AWS Sample Bedrock Invocation Data Lake GitHub repository, with an optional Amazon Quick Sight dashboard.

How TCS built an agentic test orchestrator with Amazon Bedrock AgentCore

TCS built a multi-tenant Agentic Test Orchestrator on AWS that turns Jira stories, BRDs, Swagger specs, and GitLab sources into test scenarios, Gherkin files, and executable Java or Python tests, using specialized agents hosted on Amazon Bedrock AgentCore runtime with the Strands Agents SDK. The platform separates a shared control plane from per-tenant application planes with isolated Cognito user pools, KMS keys, S3 buckets, and DynamoDB tables, and routes API tests to Amazon ECS on Fargate and Selenium UI tests to an EC2-backed ECS cluster. TCS reports that in an internal benchmark of about 100 scenarios, average effort fell from roughly two hours to about 14 minutes per scenario, an 88 percent reduction.

What a read-only Postgres role does and doesn't stop when an AI agent connects

A hands-on PostgreSQL 17.11 test shows that a carefully built read-only agent role blocks writes, DDL, COPY TO PROGRAM and ungranted tables, but its read-only and timeout settings are only per-session defaults that the agent can turn off with SET or BEGIN READ WRITE, and it can even alter its own stored defaults via ALTER ROLE. The piece also shows that GRANT SELECT exposes every column, that PUBLIC can execute SECURITY DEFINER functions like refund_order, and that pg_catalog reveals schema, roles and function source, so the real boundary is the GRANTs plus external enforcement. It closes with a nine-step checklist covering column-level grants, row-level security, revoking PUBLIC EXECUTE and default privileges, and giving the agent only its own credential.

Introducing Security Factory

Factory introduces Security Factory, now in Private Preview, which connects investigation, remediation, and verification into one continuous workflow to help engineering teams turn security findings into tested fixes. It combines Connectors for consolidating findings and context, Workstreams for parallel agent investigations, and Factory's coding workflows for developing and testing fixes. Organizations can configure agent autonomy levels, from presenting remediation plans for human approval to automatically creating pull requests and deploying fixes within approved policies.

Boxes

Amp introduces boxes: always-on computers starting at $18 per 30 days (billed by the minute) to which users can deploy the software they build with agents, with each application getting its own URL and no containers involved. Boxes can be shared with a workspace, can be turned on and off on a schedule, and each comes with its own runner so agents can operate directly on the box.

How to Build a Voice AI Agent

This tutorial builds a Python restaurant-booking voice agent that chains speech-to-text, an LLM with tool calling, and text-to-speech through a single OpenRouter API key, streaming each completed sentence to TTS instead of waiting for the full reply. Across 12 turns in three test conversations, median server-side time to first audio was 1.9 seconds (1.4 seconds without tools, 2.4 seconds with a tool call), and a four-turn conversation cost about $0.013 per minute of processed speech at October 8 promotional prices, with TTS accounting for roughly 90% of model cost. It also compares STT, LLM and TTS models on accuracy, latency and price, and shows how to run the same agent with LiveKit Agents for live audio and interruptions.

Best OCR Model for Scanned Documents, Measured

The author tested 15 vision models plus Mistral OCR and Tesseract on eight synthetic scanned pages scored by character error rate and 264 key-value checks, finding Claude Opus 5.5 most accurate (262 of 264 at $16.42 per 1,000 pages) and Mistral OCR the best value (258 of 264 at $2.02). The post documents failure modes such as fabricated card digits, wifi passwords and server names on a faded receipt, and provides code for transcription, disabling reasoning, cross-checking two models from different labs, and using the mistral-ocr engine.

Show HN: AI solved software. Now it's time to raise the bar. Introducing Brigade

Brigade is an MIT-licensed open-source product that gives you a team of AI teammates running the real Claude Code or Codex CLIs signed in with your own Claude or ChatGPT subscription accounts, each with a job, its own computer, Apps and memory. Decisions needing you land in one list on your phone, and it is free to self-host or $99 a month per workspace for Brigade Cloud. The repository includes the dashboard, API, runner, desktop app and deployment docs, and most commits come from Brigade's own teammates with a person approving every change.

How We Built TLX: A Hardware-Native GPU Compiler for Production ML at Scale

Meta introduces Triton Low-Level Extensions (TLX), an extension to Triton that adds explicit warp-group and cluster-level orchestration so engineers can use modern GPU features like TMA, TMEM, and warp specialization while keeping Triton's programming model. TLX is in production across Meta workloads, with reported gains including 20-25% end-to-end MFU for GEM model training on thousands of NVIDIA Blackwell GPUs, up to two-thirds net-latency reduction in RecSys inference, and a distributed grouped GEMM kernel running in large-scale LLM pretraining. The post frames TLX as implementing a "MIMW" abstraction between CUDA's SIMT model and Triton's block-level model, and covers five extension areas: warp-group orchestration, synchronization, memory orchestration, cluster execution, and optional layouts.

Optimizing Muse Spark Open on TPU v7 with SGLang-TorchTPU

The post describes ExecuTorch's new WebGPU backend, which lets PyTorch models exported once to a .pte file via torch.export run in any WebGPU-enabled browser through a WebAssembly runtime, covering LLMs, vision-language, segmentation, depth, detection, super-resolution, speech-to-text and on-device LoRA/DiReFT fine-tuning. On an Apple M1 Pro in Chrome, WebGPU was 4–7× faster than ExecuTorch's Wasm/XNNPACK CPU path on prefill and 2–10× faster on decode, with all data staying on the device. The backend is still in early development, and the post links an ExecuTorch WebGPU Playground and Hugging Face repos with .pte files and export recipes.

Featured coverage of this story: 《ExecuTorch WebGPU: PyTorch Models on the GPU, in the Browser》

Google broadens its AI hardware portfolio with two new specialized chips

Google unveiled two new variants of its in-house tensor processing units (TPUs) for AI workloads: TPU 8tx, designed with AMD for confidential training and inference with security built into the hardware layer, and TPU 8i Max, which combines two inference-focused TPU 8i chips in a single package to double memory capacity, processing power and bandwidth. Both are part of the company's eighth-generation chips introduced in April and broaden its portfolio to confidential inference and larger memory capacity for large workloads. The launch comes as hyperscalers and AI labs work on their own AI-specific semiconductors to reduce reliance on Nvidia.

Introducing Mastra FDE

Mastra announced Mastra FDE, a forward-deployed engineering program open to any team building agents, in which a dedicated Mastra engineer embeds with the customer's team from architecture through production at 6 hours/week consultative or 20-40 hours/week hands-on. The post cites Salesforce's Agentforce Vibes harness, MongoDB's Sage platform, and Workday moving security operations agents from Bedrock Agents to Mastra in 31 days while cutting 3,000 lines of code and about 98% of estimated vector storage costs. It positions the offering as cheaper than one senior AI hire and backed by a shared Slack channel with Mastra's founders and engineers, roadmap insights, upskilling sessions, and Mastra Enterprise.

NVIDIA Expands Open Source Commitment With OpenShell Donation and New AI Infrastructure

At KubeCon + CloudNativeCon, NVIDIA is contributing its OpenShell agent security runtime to the CNCF as a Sandbox project and contributing to Google-led Agent Substrate, also accepted into the CNCF Sandbox. NVIDIA also announced new DSX AI Factory Platform components: NVIDIA Kubernetes Engine, NVIDIA Kubernetes Recipes and NVIDIA Snapshot, which restores checkpointed replicas in seconds without model reload or warmup. Design partners Crusoe and Saturn Cloud reported reduced workload readiness and vLLM startup times using Snapshot.

Running AV Reasoning Models Within Onboard Compute and Latency Limits

AV developers need reasoning models that fit onboard compute and latency constraints; larger models provide a foundation, but deploying them in a vehicle requires reducing model size, preserving useful reasoning, and validating performance on target hardware. The source mentions NVIDIA Alpamayo 2 Edge in this context, but the supplied material is truncated before detailing it.

Show HN: Estimate the carbon footprint of your company's AI use (open data)

SOMA AI released a free calculator that estimates the CO₂e, electricity and water behind an organisation's use of AI tools such as ChatGPT, Microsoft 365 Copilot, Claude, Gemini, GitHub Copilot and AI APIs, broken down by team and tool with a range. It uses SOMA AI's published inference factors (DOI 10.5281/zenodo.23371447, method in peer review), based on ML.ENERGY Leaderboard v3 energy per token on H100 GPUs, a PUE of 1.2 and regional grid carbon intensity from Ember 2023 and US EPA eGRID 2023. A first estimate needs only tools, people and usage frequency (Tier 2b), with message counts or token data replacing defaults, and results are reported as Scope 3 Category 1 location-based.

Show HN: We wanted to build an AI application, not maintain RAG pipelines

Akropolys offers free integration of live voice, chat, and proactive personal agents directly into a website, with Kiku providing the user-facing experience. It connects to data either through batch ingestion of an array of records (Growth, Pro and Business plans) or single-record ingestion (all plans including Free), both building the same search index. A coding agent can be given the Akropolys integration skill via `npx skills add https://akropolys.cloud --skill akropolys-integration`.

Show HN: Deedglass – Verify AI Agent SQL Changes Using Postgres WAL

Deedglass is an open-source, self-hosted tool that sits between AI agents and the systems they modify: rules decide each tool call, a check compares what the target actually changed against what the agent reported (Claimed, Checked or Proven), and each result is sealed as a signed record verifiable offline. Its demo shows an agent reporting a 1-row UPDATE while a legacy trigger changed 301 rows in the same transaction, caught via Postgres logical decoding, with an undo that runs only after a person countersigns. It is pre-1.0, has one maintainer, and Proven checks cover Postgres only, requiring wal_level=logical and REPLICA IDENTITY FULL on watched tables.

New Linux-Based Desktop Operating System Is Built for AI Nerds

Kora, announced Tuesday, is a Linux-kernel desktop operating system aimed at "the AI native, not mass market," designed to integrate context across multiple AI tools at the system level and scheduled to launch sometime in 2027. The startup has partnered with Qualcomm to bring Kora to Snapdragon PCs next year, supports x86-64 architecture, and plans built-in compatibility layers for standard Windows apps alongside its own native apps and an SDK. Kora grew out of CEO Kirt McMaster's startup Playtron, whose GameOS architecture now lives inside Kora as GameFrame, and the company is optimizing for desktop configurations with at least 64 gigabytes of RAM.

Decision Models in Hugging Face

Hugging Face explains decision models, which take a state plus typed questions (choice, score, and noul, a Bernoulli yes/no primitive) and return structured decisions with probability distributions instead of generating text. The post contrasts them with traditional and zero-shot classification, and describes ecosystem support: Transformers' DecisionPipeline, AutoModelForDecision and DecisionConfig, a /v1/systemone endpoint also served by llama.cpp, a Hub tag for discovery, and fine-tuning via the Trainer API.

Worktrunk – Git worktree management for parallel AI agent workflows

Worktrunk is a CLI for Git worktree management designed for running parallel AI agent workflows, offering three core commands (switch, list, remove) that address worktrees by branch name. It adds hooks for workflow automation, LLM-generated commit messages, a one-command merge workflow, an interactive picker with CI status, and copy-on-write build cache sharing on APFS, btrfs, and XFS. It is installable via Homebrew, Cargo, Winget, Arch Linux, and Conda/Pixi.

Software correctness & supply chain security, with Sean Allen

In this interview, Antithesis distinguished engineer and Pony core contributor Sean Allen discusses software correctness, arguing that strong type systems in the ML family (such as Rust's sum types and exhaustive matching) catch errors at compile time that ALGOL-family languages miss, and that "make illegal states unrepresentable" beats manual testing. He describes "workloads" like the Bank Test for validating database isolation guarantees, pushes back on treating quality as a late CI/CD "gate," and says there is no simple fix for supply chain security because robust languages like 'E' are unpopular while popular languages assume a trustworthy ecosystem.

How to make flash models even faster: optimizing DeepSeek V4.1 Flash 1017

Baseten describes using an internal harness to optimize 80+ kernels and bottlenecks for DeepSeek V4.1 Flash 1017, producing a Fast variant it reports at 991 output tokens/s on Artificial Analysis with the lowest time per task. The post details three hardware-level techniques: kernel fusion (including an improved fused router kernel cutting decode time by 2.7ms and a 30% faster mHC megakernel), Programmatic Dependent Launch to overlap kernel execution, and host loop minimization via breakable CUDA graphs in SGLang.

Manage Kubernetes Clusters for GPU Infrastructure with NVIDIA Kubernetes Engine

NVIDIA announced two open source projects, NVIDIA Kubernetes Engine (NVKE) and NVIDIA Kubernetes Recipes (NVKR), for building and operating managed Kubernetes services on GPU infrastructure. Both are new projects available in public preview. The announcement notes that AI Factory customers often want to consume GPU capacity through the Kubernetes APIs, tools, and ecosystem they already use.

Start and Restore GPU Workloads in Seconds with NVIDIA Snapshot

NVIDIA Snapshot is a new open-source component of the NVIDIA DSX AI Factory Platform that makes initialized GPU workloads reusable across inference replicas, speeding startup and recovery. The post notes that when inference demand spikes or a replica fails, available GPUs do not immediately translate into serving capacity, since initializing an LLM workload—a cold start—can take minutes.

Agent Substrate: Advancing Kubernetes for Agents at Scale

AI agents spend most of their time idle, waiting on a human prompt, a tool result or a model response, and Kubernetes was not built for that. Running one Pod per session holds memory for inactive sessions, while deleting Pods between turns incurs a multi-second cold start and loses the agent's state, and RL post-training hits the same wall. Agent Substrate has now been donated to the foundation named in the source.

After challenging Google’s Android with Cyanogen, its co-founder takes aim at Windows for AI era

Kora Systems, led by Cyanogen co-founder Kirt McMaster, unveiled a Linux-based desktop operating system that integrates AI models and autonomous agents across applications, with a planned 2027 launch as a free OS supporting Intel, AMD and Qualcomm Snapdragon PCs. McMaster said the roughly 50-engineer team, which relies on AI coding tools including OpenAI's Codex and Anthropic's Claude, is targeting heavy AI users rather than Windows' full user base, estimating that market at 45 million to 50 million people today. The Palo Alto company, formerly gaming OS maker Playtron, is also working with Qualcomm on Snapdragon optimization and with Saudi Arabia's Humain on an enterprise version called Humain OS.

Open-Source 2026 AI Cloud GPU and Scraping Proxy Pool Cost Calculator

This interactive calculator estimates monthly costs for dedicated proxy pools (IPv6, shared IPv4, and individual IPv4 tiers) and cloud GPU compute (RTX 4090, A100, and H100) based on IP count and active hours. It displays market-average comparisons, recommended static IP counts for AI agents, and a promo code for 5% off at proxy6.net.

Almanzor – Ask AI and an MCP server for your docs, with cited answers

Almanzor Ask turns your docs or website into an assistant that answers from your own content, links the page it used, and says "I don't know" when the answer isn't there. It offers a docs MCP for coding agents such as Cursor or Claude Code, and a floating Ask button added with one line of code. It is hosted in the EU, and the company says content and conversations are never used to train models.

Show HN: ChatCairn, a local, searchable library of your AI chats

ChatCairn is a Chrome and Edge browser extension that saves conversations from ChatGPT, Claude, Gemini, Grok and Perplexity into a local, searchable library, with Markdown and PDF export and highlights turned into notes and tasks. It requires no account, keeps chats in the browser with no ChatCairn servers or analytics, and is free for up to 30 saved chats, with Pro at $4.99/month or $69 lifetime.

Hands-on with Ring’s new AI drone camera for your home

Ring says its Drone Cam, an AI-piloted indoor drone that flies pre-programmed routes to livestream and record when triggered by Ring sensors, will ship on January 27th, 2027, for $249.99. The Verge's hands-on found a 165-gram device with 2K resolution and a 148-degree field of view that relies on cloud-based AI Pilot for navigation, offers 5 to 8 minutes of flight per charge, and requires about 8 minutes of charging per minute of flight. Ring also announced the $999.99 Ring Rover Cam, and said the Drone Cam is still obtaining FCC certification.

Show HN: TeamOnly, free hosting for AI-built apps behind your company login

TeamOnly is a hosting service for internal tools built with AI coding tools such as Claude Code or Cursor, deployed with the command `npx teamonly deploy` to a private link like https://expenses.acme.teamonly.app that only coworkers on the company's email domain can open. It includes a database, file storage and secrets set up on first deploy, and sign-in via work Google or Microsoft accounts or a code sent to a work email, checked before app code runs. It offers a 14-day free trial with no card, then from $19 a month.

Show HN: pvewarden – let an AI agent patch Proxmox VMs without giving it a shell

pvewarden is an MCP server that lets an AI agent such as Claude Code check pending updates, take snapshots, install updates, reboot and verify services on Proxmox VE guests without receiving a shell or root. Safety comes from a pool-scoped Proxmox token, per-guest SSH keys bound to a forced command, required Proxmox tags for updates and reboots, automatic snapshots before installs, and confirmation dialogs the agent cannot answer. It is at version 0.1.x, tested on Proxmox VE 9 with Debian 12/13 and Ubuntu 22.04/24.04/26.04 containers and VMs, and is licensed Apache-2.0.