#Deployment and engineering

Build long-horizon agents with Grok 4.7 on Amazon Bedrock

AWS demonstrates how to build and evaluate a tool-calling agent using Grok 4.7 on Amazon Bedrock, comparing four reasoning effort levels across three synthetic tasks. The agent answered 67 of 72 runs correctly; xhigh achieved 18 of 18, while low achieved 15 of 18 using about an eighth of the output tokens. The guide covers task design, automated grading, production monitoring, and operational controls, noting that these small-scale results are indicative rather than statistically conclusive.

How ShipStation Global built an AI IT help desk on Amazon Bedrock AgentCore

ShipStation Global built Deckhand, a Slack IT help desk agent on Amazon Bedrock AgentCore, to answer questions, analyze error screenshots, and create or update Jira Service Management tickets. The implementation guide explains its Strands agent, knowledge retrieval, MCP tools, and extraction of resolved Slack threads using DSPy’s Recursive Language Model. In its first three months, 812 of 1,221 agent conversations were handled in chat, while 409 became tickets; positive feedback rose from 82 percent in April to 96 percent in June.

Simon Willison Builds web-edit 0.1a0 to Edit Files From a Phone

Simon Willison built web-edit 0.1a0, a Python web app for viewing and editing files in a folder, after wanting to edit a credentials file from his phone without sending those credentials through the LLM. He intends to run it over Tailscale so an agent can start the app and he can edit files directly from his phone. The editor uses CodeMirror 6; Willison said he had considered Monaco but found its mobile experience was not yet great.

Simon Willison Quotes Geoffrey Huntley on Enforcing Agent Use of a Diagnostics Library

Simon Willison quotes Geoffrey Huntley describing a pre-commit hook that checks the git diff for direct calls to observability modules that bypass his diagnostics library. When it finds one, the hook fails the commit and injects a prompt telling the agent to use the library. Huntley says the approach makes the agent “ten out of ten.”

Mistral Models Join an AI Satellite Constellation

Look up! Mistral models are in orbit 🌍. Proud to join @orbitworksspace and @LoftOrbital's first-of-its-kind AI satellite constellation with @NVIDIAAIInfra. @LoftOrbital says @orbitworksspace launched Altair, the first of a 50-satellite AI constellation, a few weeks ago. The constellation will enable users to receive AI-generated alerts within minutes. @NVIDIAAIInfra solutions support models like Mistral’s Ministral 3 to deliver fast, reliable answers to users' critical questions. Mistral says it is proud to join the constellation.

NVIDIA Details RTX Spark Software Updates Ahead of Laptop Launch

NVIDIA and its partners are rolling out software features and optimizations ahead of RTX Spark laptops becoming available Friday, Oct. 16. Updates include simplified local agent setup, sandboxing through OpenShell and Microsoft Execution Containers, and local and cloud model routing. NVIDIA PAIR exits beta with llama.cpp support, while Nemotron 3.5 Bolt 74B-A9B 24GB arrives Thursday, Oct. 15; creative apps, developer tools, and games also receive RTX Spark support and performance updates.

XGRIDS Reconstructs China’s Minning Town as a 6.2 Billion-Gaussian 3D Model

XGRIDS reconstructed 23.5 square kilometers of Minning Town as a continuous 3D Gaussian model containing 6.2 billion Gaussian ellipsoids, featured in a China Media Group documentary. The project used 186,000 aerial and ground images and 15 billion LiDAR points, collected over seven days; reconstruction reached centimeter-level geometric accuracy with error below 2 cm. The production team used XGRIDS’ LCC Plugin for Unreal Engine to bring the model into the documentary.

ReplaySSM reports decoding speedups and production use, with support already in vLLM

vLLM praises ReplaySSM work by @zwljohnny, @tri_dao, and @NVIDIA, saying Mamba2 standard decode and FlashInfer-backed MTP support are already in vLLM. Caching SSM inputs and updating the checkpoint only when needed reduces recurrent-state memory traffic. In the quoted post, @zwljohnny reports production use in GLM-5.3-Flash and Kimi-K3, availability in TensorRT-LLM, vLLM, FlashInfer, and SGLang, speedups of up to 1.47x for standard decoding and 2.1x for speculative decoding, and low-precision SSM state whose rounding error never propagates through the recurrence.

vLLM Support Comes to Intel’s Xe3P-Based GPU D-Series

Great work from the @IntelTech team bringing vLLM support to the new Xe3P-based Intel GPU D-series. More hardware choice for serving open models, with vLLM and the open PyTorch ecosystem.

Agile Robots Presents 12 Exhibits at the 2026 China International Industry Fair

Agile Robots brought 12 exhibits to the 26th China International Industry Fair, held in Shanghai from October 12 to 16, presenting work across five areas: embodied intelligence, embodied models, research, consumer applications and industrial deployment. Featured exhibits included the Mobile FR3 Duo research platform, an AI therapy and wellness robot, and a photography robot. The article was provided by Agile Robots and reprinted with authorization by QbitAI.

How Perplexity Projects uses Git repositories as shared workspaces for agents

Every Perplexity Project is a Git repo, built as a Git forge for agents without what Srinivas calls GitHub bloat because no human has to approve anything. Each agent gets its own short-lived sandbox and commits to a shared repo on S3, with fewer than 2% of submissions colliding. Agents search with plain `ls` and `grep`, which Srinivas reports beats semantic search 63% vs 37% on answer accuracy for financial document analysis.

The same story is shown in the picks as 《How Perplexity built the Projects filesystem as a Git forge for agents》

Guide to accessing OpenAI GPT-6 models on Amazon Bedrock with global cross-Region inference

AWS explains how to invoke OpenAI GPT-6 Sol and GPT-6 Luna through Amazon Bedrock global cross-Region inference from Thailand, Malaysia, Singapore, Indonesia, and Taiwan, and GPT-6 Astra from Singapore. The guide covers Converse, OpenAI-compatible APIs, coding clients, IAM permissions, SCP exceptions, quotas, and monitoring. Billing and quota tracking remain in the source Region, while requests run in supported destination Regions; each output token consumes 10 tokens of quota.

Simon Willison reviews Python 3.15 and adopts Python 3.11 features

Simon Willison reviews Python 3.15 and explains how Python 3.10’s end of support lets his open source projects adopt Python 3.11 features. Highlights include explicit lazy imports that address startup costs in his LLM CLI plugins, unpacking in comprehensions, and the Tachyon sampling profiler. He provides uv upgrade commands and identifies newly usable features such as tomllib, asyncio TaskGroup, and SQLite improvements, including removal of a Datasette compatibility workaround.

How Perplexity built the Projects filesystem as a Git forge for agents

This is how Perplexity built the filesystem behind Perplexity Projects as a Git forge for agents. Each agent works in its own sandbox and commits to a shared repository where people and other agents can build on its work across sessions. Read more: https://t.co/jcrTptb2IW https://t.co/dXnhroO4mg

Run multi-node inference workloads using Ray LLM Deep Learning Containers

AWS shows how to deploy Qwen3.5-9B across two GPU nodes on Amazon EKS using Ray LLM Deep Learning Containers, with separate prefill and decode deployments behind one OpenAI-compatible endpoint. The walkthrough configures the application in YAML and uses NIXL over EFA to transfer the KV cache between nodes. It covers cluster provisioning, model invocation, transport verification, and resource cleanup.

Deploy decision models with llama.cpp on Amazon SageMaker AI

AWS explains how to deploy GGUF models with its llama.cpp Deep Learning Container on Amazon SageMaker AI, then add a SemIf adapter that returns probabilities over declared options without generating answer tokens. The guide includes CPU and GPU deployment code. Across 231 JevBench tasks using Qwen3.5-4B Q4_K_M, the GPU configuration delivered 4.9 times faster p50 and 32.2 times faster p95 latency than the CPU configuration, with similar accuracy.

Aravind Srinivas Thanks @MollySOShea for the Conversation

@MollySOShea says she spoke with Perplexity Founder and CEO Aravind Srinivas about why AI-native SMBs could drive the next American millionaire economic boom. Was a fun chat! Thanks for having me @MollySOShea! https://twitter.com/mollysoshea/status/2109645765958132083

NVIDIA Opens Vera Rubin Rack Designs and Expands AI Factory Infrastructure at OCP

NVIDIA contributed Vera Rubin rack and compute tray designs to OCP through its open MGX architecture and launched open source components for DSX AI Factory. MGX 1.2 supports rack deployments up to 250kW and adds standardized power, cooling and interconnect interfaces. NVIDIA and partners also announced facility and grid resilience specifications, Meta’s adoption of Spectrum-6 switches, plans for USFP optical modules with Cisco, and Falcon Transport support on future ConnectX-10 SuperNICs and BlueField DPUs with Google.

Blueprints AI says it can generate permit-ready architectural blueprints in minutes

Blueprints AI says its software generates code-compliant, permit-ready architectural construction blueprints in minutes and will make its second-generation version publicly available this week. Version 2 accepts project descriptions or sketches and produces architecture, structural, mechanical, electrical, plumbing, and interior design blueprints, with design changes updating the entire project and being checked against local building codes. The co-founders say licensed professionals still review and seal documents before permit submission.

Microsoft plans to scale its own Maia AI chips to gigawatt levels by 2027

Microsoft plans to deploy Azure Maia AI chips at multi-gigawatt scale starting in 2027, making them its main inference system across its products, executive Scott Guthrie writes. Maia 200 already runs Microsoft Copilot in the US, Europe, and Australia, with GitHub Copilot and Foundry following by year's end. Maia 300 ships in 2027 for long inputs, fast chats, and agentic tasks; Microsoft claims 40 percent more tokens per megawatt than leading GPU systems and up to seven times the tokens per 96-chip rack compared with a Maia 200 rack.

Microsoft introduces Maia 300 with 40% more tokens per watt

Mustafa Suleyman calls Microsoft’s new silicon 🔥 and highlights “40% more tokens per watt!” In the quoted announcement, Satya Nadella says Maia 300 delivers 40% more tokens per watt than leading GPU-based systems, with multi-gigawatt Maia deployments planned starting next calendar year as the primary inference system for Microsoft’s products and services. He says Maia 200 is already being deployed across the US, Europe, and Australia, delivering up to 1.7x more tokens per watt across Teams and Copilot experiences. https://twitter.com/satyanadella/status/2109631106404487523

Qwen introduces QwenGyre for elastic RL in extreme-long-horizon agent training

Qwen introduces QwenGyre, bringing elastic RL to extreme-long-horizon (XLong) agent training. Its Elastic Scheduler reallocates GPUs between rollout and training without interrupting live executions, while its Trajectory Processor converts branching executions into prefix-shared trajectory trees with bounded, deduplicated training samples. Scaled to Qwen3.8-2.4T-A95B with ~700K tokens per rollout, QwenGyre raises NL2RepoBench from 52.5% to 58.5% in 48 steps and cuts end-to-end training time by ~44% vs Async. Paper: https://arxiv.org/abs/2609.33848

How Anta Uses Volcano Engine Agents Across Global E-Commerce Workflows

Anta is using Volcano Engine Agents for promotional videos, overseas ordering, and cross-border returns. The article reports that an ArkClaw workflow packages four ordering templates into Skills, reducing ordering decisions from half a day to 10–20 minutes, cutting manual entry by about 90%, and achieving data accuracy above 99%. A returns Agent automatically creates over 1000 return documents monthly for Anta’s TikTok cross-border business, with expansion to FILA and other brands planned.

Baidu World 2026 to Be Held in Beijing on November 12

Baidu announced that Baidu World 2026 will take place on November 12 at the China National Convention Center Phase II in Beijing, under the theme “The Next Act: Intelligent Organizations.” The event plans one main forum, 12 breakout forums, more than 30 AI classes, and a 6,000-square-meter exhibition area. Its forums will cover AI-driven organizational change, including AI office tools and enterprise deployment of agents.

Baoyu uses REA and Computer Use to investigate CapCut subtitle templates

Baoyu describes using REA with Computer Use to investigate how CapCut renders dynamic subtitles, without knowing where their formats were stored. He had it operate CapCut, download Templates, and monitor network URLs and disk save locations. After downloading several dozen through clicks, Codex found clicking too slow and constructed URLs to download directly, ultimately downloading 2291 Templates.

OpenAI raises GPT-6.1 Sol default context to 500k in Codex

OpenAI's Thibault Sottiaux said the default context length for GPT-6.1 Sol when used through Codex and Work has increased to 500k, up from 272k, an increase of about 80%. He said this reduces how often compaction occurs and helps tasks that need to retain a lot of context over time, and that users can override it, with a maximum context length of 1M.

Testing Local AI With Hermes Agent and Qwen on an M5 Ultra Mac Studio

A Verge writer began running local LLMs on an M5 Ultra Mac Studio with 256GB of unified memory, installing the open-source Hermes Agent desktop app and the 125-billion-parameter Qwen 3.8 Flash Next model, about 105GB in size. Early tasks included a daily briefing via Telegram, reorganizing a 400-plus-game Steam library, and local data analysis; the writer is now working on automating laptop benchmark tests.

Siemens to Show Industrial Full-Stack, Humanoid Robot Solution at 2026 CIIF

Siemens will showcase its industrial full-stack capabilities at the 26th China International Industry Fair, including a first full end-to-end humanoid robot solution spanning design, manufacturing, assembly and factory-floor use. It plans to release more than 20 new products and open 6 embodied intelligence "factory entry" scenarios, alongside an upgraded "钻耀之心3.0" and applications for data centers, electronics and semiconductors, and zero-carbon factories.

Silicon Valley's AI Compute Crunch Reshapes Talent and Startup Deals

A QbitAI roundup argues that compute access now drives talent retention, model leadership and monetization in Silicon Valley. It cites Noam Shazeer leaving Google for OpenAI after his team's compute was reallocated, Anthropic's May SpaceX deal for over 300 MW and 220,000 Nvidia GPUs, and SemiAnalysis data showing H100 one-year rental prices rising from $1.70 to $2.35 per GPU-hour between October 2025 and March 2026.

Python 3.15.0 added to actions/python-versions for GitHub Actions

Python 3.15.0 has been added to actions/python-versions, Simon Willison reports, meaning "3.15" can now be added to a GitHub Actions testing matrix to run tests against the new release. Willison said he had been waiting a couple of days for the update and had even asked ChatGPT to check for it.

Microsoft launches Decision-1 decision model based on Qwen3.5-9B

Microsoft released Decision-1, a decision model for classifications, evaluations and routing decisions, built on Qwen3.5-9B. Microsoft says it tops 36 benchmarks covering nearly 150,000 questions with 83.5% accuracy and 85 ms latency, and is 2.5 times faster than runner-up H2O-Lightning-4B. It is available through Microsoft Foundry and OpenRouter at $0.042 per million input tokens, with output tokens free.

Standard Bots Details Its AI Stack for Industrial Robot Arms

Standard Bots, which calls itself "America's largest AI-native industrial robot manufacturer," described its AI stack for industrial robot arms used in machine tending, welding, and assembly. Its largest model is in the low billions of parameters, trained on over a billion images for zero-shot perception, with training in the cloud and inference run on-premises via edge GPUs. The company recently raised $200 million at a $1 billion valuation in a Series C led by General Catalyst and RoboStrategy.

Leapmotor's Second Brand Aims for an iPhone-Style Revolution, Zhu Jiangming Says

Leapmotor founder Zhu Jiangming said the company's second brand aims to do to cars what the iPhone did to feature phones, with a new architecture and a cabin space efficiency above 100%. The brand's first model is planned to debut and begin deliveries in Q4 next year. Leapmotor also detailed a world-model smart driving stack and technology supply deals with FAW and Stellantis.

Tesla Renames FSD to Assisted Driving in Europe

Tesla has renamed Full Self-Driving (Supervised) to Tesla Assisted Driving on its websites in Germany, Spain, the Netherlands and Slovakia, while the US site keeps the original name. Germany's transport ministry called the FSD name misleading, and Tesla proposed the change itself. Germany's transport minister said he would push for EU-wide approval, though a planned October vote was delayed to at least December.

GMIF 2026: 140 trillion Token calls per day are changing SSD capabilities and division of labor

As of March this year, China's daily average Token call volume exceeded 140 trillion, and large-scale inference is driving SSDs to extend from capacity and persistent storage toward the compute side. Samsung demonstrated a Z-NAND architecture, placing model weights in Z-NAND and keeping KV Cache in DRAM; large North American internet companies have begun replacing some HDDs with high-capacity QLC SSDs. Data management is also shifting from fallback-style tiering to proactive tiering at the outset of system design and dynamic scheduling during operation.

vLLM adds support for NVIDIA Vera Rubin

vLLM now supports NVIDIA Vera Rubin, with early results showing more than 7.8x the throughput of GB200 on MiniMax M3 on AgentX. The vLLM team says it has been bringing vLLM up on Rubin with Inferact, NVIDIA AI, Red Hat AI and the vLLM community since Rubin was announced.

Amazon and others are no longer keeping data center deals secret — is that enough to build trust?

Amazon announced it will no longer use non-disclosure agreements when negotiating data center deals with local governments, after Microsoft took a similar step earlier this year. The secrecy practices had previously fueled community opposition to AI infrastructure, with hundreds of proposed and enacted bans from New York to San Francisco. TechCrunch's Equity podcast discussed whether dropping NDAs will change how data center deals get done, and why trust may be the real product for personal AI startups.

AWS September Roundup: An Overview of Amazon Bedrock, AgentCore, and Strands Updates

Amazon Bedrock has launched OpenAI's GPT-6 Astra, Sol, and Luna series models, among which GPT-6 Astra supports up to 1 million input tokens, and the Ultrafast version boosts API inference speed by up to 6x, reaching 300 tokens/second.

How Ai2 Rebuilt Cluster Scheduling with GPU Time Budgets

Ai2 replaced its original priority scheduler with GPU time budgets, hierarchical fair allocation, and time-slice contracts, turning debates over how much GPU time each research project should get from case-by-case operational decisions into a transparent budget allocation process. Its cluster manages thousands of NVIDIA H100, B200, and B300 GPUs, with scales ranging from 88 to 1,024 GPUs, and demand from roughly 150 internal researchers has long been 2-3 times the available compute.

How Postman runs Agent Mode on Amazon Bedrock for 40 million developers

Postman runs Agent Mode on Amazon Bedrock, providing 40 million developers with an AI-native way of working that spans API testing, documentation, discovery, and implementation.