MiniMax announced that MiniMax M3.1 Flash is now available on Tenstorrent Cloud, can run on Tensix chips, and supports Ethernet scale-out and an open-source technology stack. Tenstorrent's quoted post stated that MiniMax released the open weights of this model on the same day, and users can try it in the playground or connect it to their own agent harness.
Matched fieldsSummary
dex announced that the final version of the Software Factory Masterclass with @vaibcode and @zachlloydtweets is now live, covering the building blocks of a software factory, metrics-driven self-improvement, turning successful agent runs into evals to test future models and harnesses, as well as DORA, why counting by lines of code is a bad metric, and why counting by number of PRs may be sufficient.
Matched fieldsSummary、Body
Artificial Analysis published preliminary test results on AA-AgentPerf for Meta's new open-weight model Muse Spark Open, an 816B-parameter, 42B-active MoE that scores 39 on the Artificial Analysis Intelligence Index. Tests show NVIDIA B300 and GB300 each reach about 150M tokens/s per megawatt under throughput-first configurations, with GB300 at about 108M and B300 at about 92M at 100 tok/s output speed; AMD MI355X reaches 70M tokens/s per megawatt at 65 tok/s. 46 of the model's 62 layers use 2,048-token sliding window attention, reducing KV cache requirements at 1M context by about 75% versus full attention; Meta released an NVFP4 checkpoint and AMD released an MXFP4 checkpoint.
Artificial Analysis published preliminary test results on AA-AgentPerf for Meta's new open-weight model Muse Spark Open, an 816B-parameter, 42B-active MoE that scores 39 on the Artificial Analysis Intelligence Index. Tests show NVIDIA B300 and GB300 each reach about 150M tokens/s per megawatt under throughput-first configurations, with GB300 at about 108M and B300 at about 92M at 100 tok/s output speed; AMD MI355X reaches 70M tokens/s per megawatt at 65 tok/s. 46 of the model's 62 layers use 2,0
Matched fieldsSummary、Body
AWS has published a practical guide for multi-turn reinforcement learning (MTRL) on Amazon SageMaker AI. Using Wordle as an example, it deploys the agent and reward function to Amazon Bedrock AgentCore, and the training job updates LoRA adapters via GRPO. In the example, gpt-oss-20b at default reasoning effort improved its solve rate on held-out words from 61% to 84%, at a training token cost of about $339 and an AgentCore runtime cost of about $79. The article also details failures such as two inverted reward design errors, token invalidation under concurrency, and tool calls degrading into plain text, and recommends evaluating under the decoding settings used in final serving.
AWS has published a practical guide for multi-turn reinforcement learning (MTRL) on Amazon SageMaker AI. Using Wordle as an example, it deploys the agent and reward function to Amazon Bedrock AgentCore, and the training job updates LoRA adapters via GRPO. In the example, gpt-oss-20b at default reasoning effort improved its solve rate on held-out words from 61% to 84%, at a training token cost of about $339 and an AgentCore runtime cost of about $79. The article also details failures such as two
Matched fieldsSummary、Body
Figma announced that code layers have entered public beta, allowing teams to explore and build code-backed prototypes on the canvas. The article lists how six teams—Superside, Dell, Cox Automotive, Rocket Money, Merkle, and Checkout.com—use it, including using the Figma agent to turn static designs into code layers, comparing multiple versions side by side, and iterating in stages with precise values and reference screenshots. Merkle says iteration and feature development speed increased by 30%, and documentation creation time was reduced by up to 50%; a product manager at Cox Automotive says 19 versions took only about an hour of manual operation time. Code layers will consume AI credits; creating and editing will be free on paid plans until November 4, 2026.
ild code-backed prototypes on the canvas. The article lists how six teams—Superside, Dell, Cox Automotive, Rocket Money, Merkle, and Checkout.com—use it, including using the Figma agent to turn static designs into code layers, comparing multiple versions side by side, and iterating in stages with precise values and reference screenshots. Merkle says iteration and feature development speed increased by 30%, and documentation creation time was reduced by up to 50%; a product manager at Cox Automot
Matched fieldsSummary、Body
Italy's largest real estate platform Immobiliare.it, based on ElevenAgents, scaled from a single conversational agent to an interconnected agent network, increasing lead qualification conversion rate by 3.3x, raising the likelihood of users leaving a phone number for follow-up by 74%, and about 80% of users rated the feature positively. The first agent was deployed on listing pages, responding 24/7 in the local language and summarizing each interaction; subsequently it expanded to agents for private homeowner outreach, listing quality outbound calls, voice and chat customer service, and conversational search, orchestrated with sub-agent division of labor.
Italy's largest real estate platform Immobiliare.it, based on ElevenAgents, scaled from a single conversational agent to an interconnected agent network, increasing lead qualification conversion rate by 3.3x, raising the likelihood of users leaving a phone number for follow-up by 74%, and about 80% of users rated the feature positively. The first agent was deployed on listing pages, responding 24/7 in the local language and summarizing each interaction; subsequently it expanded to agents for pri
Matched fieldsSummary、Body
The ModelBest MiniCPM Spark Program S1 Challenge has wrapped up, with 16 projects from developers in China, Spain, Italy, Australia, and the United States winning awards. Three special awards went to Dubedo, ESP32-S3, and a personal Admin Agent evaluation project, each receiving a 1-year Typeless membership. The winning projects cover practical applications, deployment tools, and training research.
Matched fieldsSummary
At the 26th China International Industry Fair, JAKA Robotics globally debuted the Agent-driven JAKA Zu50 heavy-payload collaborative robot and the JAKA Kargo50 heavy-duty wheeled humanoid robot, which can autonomously transport material boxes of varying sizes and appearances. The site also demonstrated the JAKA K series dual-arm humanoid robot collecting process data and feeding it back to the JAKA EVO base to iterate skill models in a data flywheel. According to MIR Databank data, in 2024 JAKA's collaborative robots reached a 21.9% domestic market share, ranking first in China, and about 8.2% globally, with cumulative deployments of nearly 30,000 robots.
At the 26th China International Industry Fair, JAKA Robotics globally debuted the Agent-driven JAKA Zu50 heavy-payload collaborative robot and the JAKA Kargo50 heavy-duty wheeled humanoid robot, which can autonomously transport material boxes of varying sizes and appearances. The site also demonstrated the JAKA K series dual-arm humanoid robot collecting process data and feeding it back to the JAKA EVO base to iterate skill models in a data flywheel. According to MIR Databank data, in 2024 JAKA'
Matched fieldsSummary、Body
TinyFish released two technical guides: one defines a production-grade agent search architecture, and the other summarizes lessons from building a deep research system. The architecture is built on four web primitives—TinySearch, TinyFetch, TinyBrowser, and TinyAgent—and runs a seven-stage control loop around a foundation model, covering task definition, retrieval planning, tool routing, evidence evaluation, and failure retry. In OpenBenchmarks tests with a fixed gpt-5.6-sol model and 100 tasks, TinyFish search plus fetch achieved a 79.0% completion rate, with a median task cost of $0.050 and a runtime of 24 seconds; under a search-only configuration, the completion rate was 59.3%.
TinyFish released two technical guides: one defines a production-grade agent search architecture, and the other summarizes lessons from building a deep research system. The architecture is built on four web primitives—TinySearch, TinyFetch, TinyBrowser, and TinyAgent—and runs a seven-stage control loop around a foundation model, covering task definition, retrieval planning, tool routing, evidence evaluation, and failure retry. In OpenBenchmarks tests with a fixed gpt-5.6-sol model and 100 tasks,
Matched fieldsSummary、Body
MiniMax has publicly released the MiniMax M3.1 Flash weights on Hugging Face; the model has 428B total parameters and 23B activated per token. The company claims generational improvements in agent tasks, coding, and cybersecurity, with scores of 76.2 on BankerToolBench, 73.7% on DeepSWE v1.1, 61.7 on OSWorld V2, 29.6 on Agents' Last Exam, and 82.2 on CyberGym. This model is the stealth model Space Bunny on OpenRouter, released on October 19, and is now available on the MiniMax platform and API.
Matched fieldsSummary
robots.txt is only a crawling permission declaration, not access control; it can only constrain which paths compliant crawlers may fetch, and cannot control non-compliant crawlers, content that has already leaked, or acquisition methods that do not go through crawlers. Following RFC 9309, the article explains rules such as User-agent, Allow, Disallow, and the 500 KiB read limit, and provides robots.txt syntax for blocking GPTBot, CCBot, ClaudeBot, and Bytespider by UA, as well as methods for using nginx to apply differentiated rate limiting by UA, using logs to count request volumes per crawler, and using curl to verify whether rate limiting takes effect. The author emphasizes that robots.txt expresses intent, server-side rate limiting controls costs, logs reveal the facts, and licensing agreements address legal issues—each of the four handles a distinct part.
ontrol non-compliant crawlers, content that has already leaked, or acquisition methods that do not go through crawlers. Following RFC 9309, the article explains rules such as User-agent, Allow, Disallow, and the 500 KiB read limit, and provides robots.txt syntax for blocking GPTBot, CCBot, ClaudeBot, and Bytespider by UA, as well as methods for using nginx to apply differentiated rate limiting by UA, using logs to count request volumes per crawler, and using curl to verify whether rate limiting
Matched fieldsSummary
Matt Pocock adds a "software-factory" design Skill to the 301K ⭐️ Skills repository; it is a glossary that requires users to invoke it explicitly (disable-model-invocation: true), used for designing remote, stateless agent software factories. Core terms include Task, Trigger, Run, Runner, and service slots such as Ticket tracker, Git host, and Orchestrator, along with four slot research documents and an example task chain. The design approach emphasizes Start slow: every trigger is an accelerator, and must be paired with three brakes—automated checks, automated review, and human review—plus a loop limit.
design Skill to the 301K ⭐️ Skills repository; it is a glossary that requires users to invoke it explicitly (disable-model-invocation: true), used for designing remote, stateless agent software factories. Core terms include Task, Trigger, Run, Runner, and service slots such as Ticket tracker, Git host, and Orchestrator, along with four slot research documents and an example task chain. The design approach emphasizes Start slow: every trigger is an accelerator, and must be paired with three brak
Matched fieldsSummary、Body
Google CEO Pichai revealed in the "Sources" interview that Gemini 4 Argon processed about 25 trillion tokens in 72 hours of launch, 3 times that of Gemini 3.0 in the same period, and is still rate-limited due to compute constraints. He explained for the first time the reason for cutting Gemini 3.5 Pro: the team found a gap in the model on Agentic Coding, first used the Flash version to complete intensive iterations from 3.5 to 3.8 in about 3 months, while Argon pretraining had already been completed at that time, so they injected the accumulated gains into Argon post-training. He also mentioned that Google internally has entered the RSI stage of researchers and models collaborating, and said the next-generation Gemini 5.0 will be much stronger than Argon.
the same period, and is still rate-limited due to compute constraints. He explained for the first time the reason for cutting Gemini 3.5 Pro: the team found a gap in the model on Agentic Coding, first used the Flash version to complete intensive iterations from 3.5 to 3.8 in about 3 months, while Argon pretraining had already been completed at that time, so they injected the accumulated gains into Argon post-training. He also mentioned that Google internally has entered the RSI stage of researc
Matched fieldsSummary、Body
JFrog's AI Security suite extends shadow AI detection from code repositories and builds to developer computers, deployed via Agent Plugin through MDM or Agent-Harness settings, scanning Skills, Plug-ins, and MCP-Server, returning only metadata and risk assessments. It also adds BYOK semantic AI Asset Scanning, with prompts and Skills sent directly to the customer's own model endpoints, initially supporting Claude Opus 4.8, GPT-5.6 Sol, and Gemini 3.1 Pro. Both features are now available to JFrog Unified Security and Ultimate Security customers.
JFrog's AI Security suite extends shadow AI detection from code repositories and builds to developer computers, deployed via Agent Plugin through MDM or Agent-Harness settings, scanning Skills, Plug-ins, and MCP-Server, returning only metadata and risk assessments. It also adds BYOK semantic AI Asset Scanning, with prompts and Skills sent directly to the customer's own model endpoints, initially supporting Claude Opus 4.8, GPT-5.6 Sol, and Gemini 3.1 Pro. Both features are now available to JFrog
Matched fieldsSummary、Body
Meta CEO Zuckerberg said on the October 21 earnings call that since launching on September 8, over 10 million users in the US and Canada have created Muse agents, roughly more than double Threads at its start. Muse can browse the web, send emails, and book travel on users' behalf; a quarter of users use it daily, the most common scenario is shopping, average purchase exceeds $200, first-week transaction volume exceeded expectations, and Meta had to work with Stripe to raise credit limits. Meta plans to charge merchants when Muse helps them complete sales; consumer use is free unless more tokens or more advanced models are needed; the CFO said it is working with Shopify, Walmart, and others to develop the Personal Agent Protocol.
Meta CEO Zuckerberg said on the October 21 earnings call that since launching on September 8, over 10 million users in the US and Canada have created Muse agents, roughly more than double Threads at its start. Muse can browse the web, send emails, and book travel on users' behalf; a quarter of users use it daily, the most common scenario is shopping, average purchase exceeds $200, first-week transaction volume exceeded expectations, and Meta had to work with Stripe to raise credit limits. Meta p
Matched fieldsSummary、Body
AgiBot releases the one-stop full-process embodied intelligence operation platform Genie Studio 2.0, with three major modules—Flywheel, Agent, and Monitor—covering the full lifecycle of collection-training-testing-inference-programming-management, enabling deployment validation of new operation scenarios from 0 to 1 in as fast as 7 days. The platform has been implemented in scenarios such as Changzhou CNC machine tools, Junpei Intelligent BMS assembly production line, Longqi Technology tablet loading/unloading production line, and Guangzhou Metro, among which the BMS assembly task success rate remains stable above 99.9%, the single operation cycle is controlled within 13 seconds, and Guangzhou Metro already supports the deployment of more than 60 devices.
AgiBot releases the one-stop full-process embodied intelligence operation platform Genie Studio 2.0, with three major modules—Flywheel, Agent, and Monitor—covering the full lifecycle of collection-training-testing-inference-programming-management, enabling deployment validation of new operation scenarios from 0 to 1 in as fast as 7 days. The platform has been implemented in scenarios such as Changzhou CNC machine tools, Junpei Intelligent BMS assembly production line, Longqi Technology tablet lo
Matched fieldsSummary、Body
Claude Models on Google Cloud (Anthropic managed) is now generally available, and Google Cloud customers get the full set of Claude API features, with new features and betas launching the same day as on the native Claude Platform. The offering covers Claude Managed Agents, Code Execution, Computer Use, Agent Skills, and Advisor Strategy, and supports Google Cloud IAM credentials, billing, and commitment retirement.
Matched fieldsSummary、Body
MiniMax M3.1 Flash weights are now public, with 428B total parameters and 23B activated per token. The company claims substantial improvements in agentic tasks, coding, and cybersecurity: BankerToolBench 76.2, DeepSWE v1.1 73.7%, OSWorld V2 61.7, Agents' Last Exam 29.6, CyberGym 82.2. The model uses MiniMax Sparse Attention and DSpark speculative decoding to improve inference efficiency; the weights are now available on Hugging Face and ModelScope, with API and MCode desktop entry points also provided.
MiniMax M3.1 Flash weights are now public, with 428B total parameters and 23B activated per token. The company claims substantial improvements in agentic tasks, coding, and cybersecurity: BankerToolBench 76.2, DeepSWE v1.1 73.7%, OSWorld V2 61.7, Agents' Last Exam 29.6, CyberGym 82.2. The model uses MiniMax Sparse Attention and DSpark speculative decoding to improve inference efficiency; the weights are now available on Hugging Face and ModelScope, with API and MCode desktop entry points also pr
Matched fieldsSummary、Body
Artificial Analysis released Coding Agent Index v2.0, with Claude Code paired with Claude Opus 5.5 (max) topping at 61.7 points, followed by Claude Sonnet 5.5 (max) at 59.3 points. The new version replaces DeepSWE with Cognition's FrontierCode v1.1; Opus 5.5 scores 56.0% on that benchmark, ahead of Codex paired with DeepSeek V4.1 Flash at 52.6%. SWE-Atlas-QnA-AA switched to GPT-6 Luna (xhigh) for judging, with five votes per item; the average score across 21 configurations fell from 58.8% to 57.1%.
Artificial Analysis released Coding Agent Index v2.0, with Claude Code paired with Claude Opus 5.5 (max) topping at 61.7 points, followed by Claude Sonnet 5.5 (max) at 59.3 points. The new version replaces DeepSWE with Cognition's FrontierCode v1.1; Opus 5.5 scores 56.0% on that benchmark, ahead of Codex paired with DeepSeek V4.1 Flash at 52.6%. SWE-Atlas-QnA-AA switched to GPT-6 Luna (xhigh) for judging, with five votes per item; the average score across 21 configurations fell from 58.8% to 57.
Matched fieldsSummary
The lowest price for the Alibaba Cloud Model Studio Token Plan personal edition is 39 yuan/month, corresponding to the Lite plan (original price 60 yuan/month, limited-time discounted price 39 yuan/month). It includes 11,500 Credits per month and supports 1-2 concurrent Agents. This plan covers models for text generation, image generation, video generation, speech recognition, and real-time voice conversation. The quota is calculated by subscription month and does not roll over. After it is used up, you can purchase a usage pack for 100 yuan/month that includes 20,000 Credits (up to 5 can be held at the same time). Currently, only the China North 2 (Beijing) region is supported.
month, corresponding to the Lite plan (original price 60 yuan/month, limited-time discounted price 39 yuan/month). It includes 11,500 Credits per month and supports 1-2 concurrent Agents. This plan covers models for text generation, image generation, video generation, speech recognition, and real-time voice conversation. The quota is calculated by subscription month and does not roll over. After it is used up, you can purchase a usage pack for 100 yuan/month that includes 20,000 Credits (up to 5
Matched fieldsSummary