←Search results

By relevance · Total 2779 results Clear search

Friday

  1. ElevenLabs Blog

    Warm Transfer vs. Cold Transfer: How to Choose the Right Call Transfer Method

    The ElevenLabs blog compared the differences between warm transfer and cold transfer in call centers: warm transfer provides the receiving party with background information about the caller before connecting, while cold transfer connects directly without prior notice. The article recommends warm transfer for complex or sensitive issues and cold transfer for simple, high-frequency routing, and notes that ElevenAgents supports three transfer types, with Conference Transfer on its native Twilio integration able to pass warm transfer information to human agents.

    Warm, tepid, and cold transfer comparison; operator explains warm, caller explains cold.

    The ElevenLabs blog compared the differences between warm transfer and cold transfer in call centers: warm transfer provides the receiving party with background information about the caller before connecting, while cold transfer connects directly without prior notice. The article recommends warm transfer for complex or sensitive issues and cold transfer for simple, high-frequency routing, and notes that ElevenAgents supports three transfer types, with Conference Transfer on its native Twilio int

    Matched fieldsBody

  2. Baseten Base Labs: Model Research

    Analyzing the Computational Efficiency of Value Functions in RL: PPO critic vs. GRPO | Base Labs

    Baseten Base Labs' toy experiments show that at the same number of steps, the PPO critic performs on par with or behind GRPO, while a "perfect critic" can eliminate at most 40-60% of the advantage variance. The authors argue that a value function is only worthwhile when the gradient noise it eliminates exceeds its training cost; the only scenario where it might help is with a small value head, but even then the gains are only marginal.

    Harry

    Matched fieldsBody

  3. Zhidongxi

    D-Robotics and Unitree Robotics Jointly Launch Embodied Intelligence Education Solution

    D-Robotics and Unitree Robotics made their first joint appearance at the Changsha Educational Equipment Exhibition and announced a strategic partnership, launching an embodied intelligence education solution with the Sunrise S100 compute box as the core compute foundation, combined with Unitree's Go2 quadruped robot and R1 humanoid robot. The solution comes with 15 themed courses totaling 180 lessons, covering elementary, middle, and high school, and includes 6 course-competition integrated courses. The two parties take education as their first implementation scenario and plan to further expand cooperation to industrial and other scenarios.

    D-Robotics and Unitree Robotics made their first joint appearance at the Changsha Educational Equipment Exhibition and announced a strategic partnership, launching an embodied intelligence education solution with the Sunrise S100 compute box as the core compute foundation, combined with Unitree's Go2 quadruped robot and R1 humanoid robot. The solution comes with 15 themed courses totaling 180 lessons, covering elementary, middle, and high school, and includes 6 course-competition integrated cour

    Matched fieldsBody

  4. Arena

    Arena analyzes 26,225 Agent sessions: without caching, agent costs are 3.6 times higher

    Arena analyzed 26,225 Agent sessions, 30 agents, and 13 labs from September 12 to October 23, 2026, and found that if billed at the full input price, the typical task cost for the top ten agents would be 3.6 times higher, and 592% higher on MiMo V2.6 Pro. For agents, 86% to 96% of tokens are repeatedly read context, while their own output accounts for only 1% to 2%. Cache invalidation increases the typical task cost by about 5%, and on Claude Fable 5.1, 1 in every 10 sessions sees an increase of more than 18%.

    Arena analyzed 26,225 Agent sessions, 30 agents, and 13 labs from September 12 to October 23, 2026, and found that if billed at the full input price, the typical task cost for the top ten agents would be 3.6 times higher, and 592% higher on MiMo V2.6 Pro. For agents, 86% to 96% of tokens are repeatedly read context, while their own output accounts for only 1% to 2%. Cache invalidation increases the typical task cost by about 5%, and on Claude Fable 5.1, 1 in every 10 sessions sees an increase of

    Matched fieldsBody

  5. Cohere Product and Research Blog

    Cohere Advances CAD 4 Billion AI Investment in Canada, Will Create 1,500+ Jobs

    Cohere has announced progress on its investment plan in Canada, expecting to invest more than CAD 4 billion in R&D and AI infrastructure by 2031 and create more than 1,500 jobs, of which about 1,000 will be technical roles in AI research, engineering, and product development. The plan began in 2024 and will also provide more than 250 paid student positions by 2029, along with research partnerships with the University of Toronto, the University of Waterloo, and others. Cohere has reached a CAD 309 million GPU cloud partnership with Bell AI Fabric to run its models on Canadian sovereign infrastructure.

    Cohere logo

    Cohere has announced progress on its investment plan in Canada, expecting to invest more than CAD 4 billion in R&D and AI infrastructure by 2031 and create more than 1,500 jobs, of which about 1,000 will be technical roles in AI research, engineering, and product development. The plan began in 2024 and will also provide more than 250 paid student positions by 2029, along with research partnerships with the University of Toronto, the University of Waterloo, and others. Cohere has reached a CAD 30

    Matched fieldsBody

  6. ElevenLabs Blog

    How TELUS Digital Used ElevenAgents Outbound Calling Agents to Cut Early Churn by More Than Half

    TELUS Digital built an outbound voice agent based on ElevenAgents that proactively calls customers who have just activated home internet, resulting in a 30-day churn rate for customers who received the call that is less than half that of new customers overall. The agent can disclose its AI identity, explain the first bill, guide customers to self-service tools, and remind them of bundle discounts, and can authenticate customers during the call via SMS verification code and read real-time billing data through an API. TELUS Digital selected ElevenAgents after evaluating options from November 2025 to February 2026 against five criteria: voice expressiveness, agent framework flexibility, real-time emotional adjustment, enterprise security and compliance, and response latency.

    Close-up portrait of a red-haired man with dark circles under his eyes.

    TELUS Digital built an outbound voice agent based on ElevenAgents that proactively calls customers who have just activated home internet, resulting in a 30-day churn rate for customers who received the call that is less than half that of new customers overall. The agent can disclose its AI identity, explain the first bill, guide customers to self-service tools, and remind them of bundle discounts, and can authenticate customers during the call via SMS verification code and read real-time billing

    Matched fieldsBody

  7. Leiphone

    iFlytek Chairman and President Voluntarily Cut Salaries, Each Forgoing Over 20 Million Yuan; Baidu Confirms Wu Tian's Departure, Wenxin Model to Announce New Progress Next Week

    iFlytek Responds to Executive Compensation Questions: Chairman Liu Qingfeng and President Wu Xiaoru Voluntarily Lowered Their Annual Salaries, Actually Receiving 4.1 Million Yuan and 3 Million Yuan Respectively for 2025, With Both Seeing Their Payable Compensation Reduced by Over 20 Million Yuan. Baidu Confirms That Group Vice President and Baidu Research Deputy Director Wu Tian Has Departed, and the Wenxin Model Will Announce New Progress Publicly Next Week. The U.S. Department of Justice Has Launched an Antitrust Investigation Into Nvidia's $12.9 Billion Acquisition of Hugging Face, Focusing on GPU Bundling and Control of AI Ecosystem Distribution Nodes.

    iFlytek Responds to Executive Compensation Questions: Chairman Liu Qingfeng and President Wu Xiaoru Voluntarily Lowered Their Annual Salaries, Actually Receiving 4.1 Million Yuan and 3 Million Yuan Respectively for 2025, With Both Seeing Their Payable Compensation Reduced by Over 20 Million Yuan. Baidu Confirms That Group Vice President and Baidu Research Deputy Director Wu Tian Has Departed, and the Wenxin Model Will Announce New Progress Publicly Next Week. The U.S. Department of Justice Has L

    Matched fieldsBody

  8. 36Kr · AI

    Unitree Technology and Sequoia China Establish Shanyu·Innovation Program, Entering Early-Stage Investment in Embodied Intelligence

    Unitree Technology announced on October 28 the joint establishment of the Shanyu·Innovation Program with Sequoia China, focusing on investment in embodied intelligence, robotics applications, and upstream and downstream core technologies, and proposed five principles: understanding technology and industry, discovering early, deciding quickly, valuing people, and not causing disruption. Unitree's previously launched Yuchuang Program received more than 200 projects in just over a month, and the company hopes to incubate more than 200 ecosystem enterprises over the next five years. Financial results disclosed the same day show that revenue for the first three quarters was 1.817 billion yuan, up 55.71% year on year; net profit attributable to shareholders was 387 million yuan, up 270.54% year on year; and net profit for the third quarter alone was 113 million yuan, down 17.37% year on year.

    Unitree Technology announced on October 28 the joint establishment of the Shanyu·Innovation Program with Sequoia China, focusing on investment in embodied intelligence, robotics applications, and upstream and downstream core technologies, and proposed five principles: understanding technology and industry, discovering early, deciding quickly, valuing people, and not causing disruption. Unitree's previously launched Yuchuang Program received more than 200 projects in just over a month, and the co

    Matched fieldsBody

  9. 36Kr · AI

    MaaS Has Arrived: Anthropic Turns an Operating Profit for the Quarter, China's Public Cloud MaaS Revenue Reaches 10.63 Billion Yuan in Half a Year

    Anthropic's revenue in the second quarter of this year was 11.555 billion US dollars, achieving an operating profit of 974 million US dollars for the first time, with about 80% of revenue coming from pay-as-you-go billing. IDC data shows that in the first half of 2026, China's public cloud MaaS market revenue reached 10.63 billion yuan, up 889.4% year on year, and its share of public cloud PaaS revenue rose from 5.2% in the second half of last year to 21.0%. Zhipu's open platform and API revenue was 825 million yuan, accounting for 86.5% of total revenue; Volcano Engine ranked first with a 45.1% share. The article argues that model services can now generate gross profit, and the unit of competition has shifted to inference optimization and delivery capability.

    Anthropic's revenue in the second quarter of this year was 11.555 billion US dollars, achieving an operating profit of 974 million US dollars for the first time, with about 80% of revenue coming from pay-as-you-go billing. IDC data shows that in the first half of 2026, China's public cloud MaaS market revenue reached 10.63 billion yuan, up 889.4% year on year, and its share of public cloud PaaS revenue rose from 5.2% in the second half of last year to 21.0%. Zhipu's open platform and API revenue

    Matched fieldsBody

  10. 36Kr · AI

    OpenAI Updates Model Spec, Defining Permissions and Shutdown Rules for Agents

    On October 29, OpenAI updated the Model Spec, upgrading the model behavior guidelines around five values: being loyal to human intent, reducing harm, honesty and transparency, promoting human autonomy, and caring for others. The new version shifts the focus from what chatbots can say to how much agents can do, stipulating that agents must act within agreed autonomy boundaries, remain correctable, and cooperate with shutdown, while establishing a six-level instruction hierarchy of Root, System, Developer, User, Guideline, and default no-permission. OpenAI stated that models currently running on ChatGPT and the API have not yet fully met the requirements of this specification.

    On October 29, OpenAI updated the Model Spec, upgrading the model behavior guidelines around five values: being loyal to human intent, reducing harm, honesty and transparency, promoting human autonomy, and caring for others. The new version shifts the focus from what chatbots can say to how much agents can do, stipulating that agents must act within agreed autonomy boundaries, remain correctable, and cooperate with shutdown, while establishing a six-level instruction hierarchy of Root, System, D

    Matched fieldsBody

  11. 36Kr · AI

    AI data centers are crowding out memory capacity, and PC and smartphone shipments both fell sharply in the third quarter of 2026

    In the third quarter of 2026, global PC shipments fell 20.1% year on year to 62.7 million units, and smartphone shipments fell 22.0% year on year to 255.3 million units; IDC said both were the largest declines on record. The article argues that the cause is that AI servers give priority to occupying HBM, server DRAM, and enterprise SSD capacity, leading to a shortage of consumer-grade memory; the cost share of DRAM and SSDs in PCs rose from about 15% to about 40%, and smartphone memory costs rose about 350% in a year. In the same period, monthly global semiconductor market shipments reached $159.7 billion, up 144.3% year on year, with growth coming mainly from AI GPUs, HBM, and memory price increases rather than end-device shipments.

    In the third quarter of 2026, global PC shipments fell 20.1% year on year to 62.7 million units, and smartphone shipments fell 22.0% year on year to 255.3 million units; IDC said both were the largest declines on record. The article argues that the cause is that AI servers give priority to occupying HBM, server DRAM, and enterprise SSD capacity, leading to a shortage of consumer-grade memory; the cost share of DRAM and SSDs in PCs rose from about 15% to about 40%, and smartphone memory costs ros

    Matched fieldsBody

  12. TMTPost · AGI

    SpaceX and Tesla Study Potential Merger, Musk Consolidates AI Compute Footprint

    Bloomberg reported on October 27 that Musk's SpaceX and Tesla are studying a potential merger through a special committee, and the combined market value would exceed $3 trillion once the deal closes. The article says the two companies already overlap on the Terafab chip project; SpaceX's AI business unit SI posted third-quarter revenue of $7.56 billion, up 830% year over year, with computing capacity of 1.7 gigawatts as of the end of September. The article also notes that SpaceX's Grok 4.7 scored 37.6% on the Terminal-Bench leaderboard, trailing Opus 5 with Claude Code at 53.9% and GPT-6 Sol with Codex at 49.4%.

    Bloomberg reported on October 27 that Musk's SpaceX and Tesla are studying a potential merger through a special committee, and the combined market value would exceed $3 trillion once the deal closes. The article says the two companies already overlap on the Terafab chip project; SpaceX's AI business unit SI posted third-quarter revenue of $7.56 billion, up 830% year over year, with computing capacity of 1.7 gigawatts as of the end of September. The article also notes that SpaceX's Grok 4.7 score

    Matched fieldsBody

  13. TMTPost · AGI

    DeepSeek's second funding round raises 105.5 billion yuan, valuation reaches 615.5 billion yuan

    According to foreign media citing sources, DeepSeek's second funding round raised 105.5 billion yuan (equivalent to 15.8 billion US dollars), with nearly 30 investors contributing jointly, at a valuation of 615.5 billion yuan. Industrial investors include ChangXin Memory Technologies, SMIC, Tencent, Meituan, CATL and others, each investing 1 billion to 3 billion yuan. The original article also noted that in the first funding round, investors other than the National AI Fund were required to give up voting rights and lock up their shares for five years, giving Liang Wenfeng 100% of voting rights.

    According to foreign media citing sources, DeepSeek's second funding round raised 105.5 billion yuan (equivalent to 15.8 billion US dollars), with nearly 30 investors contributing jointly, at a valuation of 615.5 billion yuan. Industrial investors include ChangXin Memory Technologies, SMIC, Tencent, Meituan, CATL and others, each investing 1 billion to 3 billion yuan. The original article also noted that in the first funding round, investors other than the National AI Fund were required to give

    Matched fieldsBody

Saturday

  1. Databricks Blog

    Wealth Management Client Onboarding Needs More Than AI: Databricks on Governed Case Context

    Databricks points out that the bottleneck in wealth management client onboarding is not any single task, but rather that evidence, case status, and decisions are scattered across CRM, KYC, screening, custody, and workflow systems. Its Data + AI Platform uses governed case context to provide AI agents and human reviewers with a shared, versioned, and traceable view, with a supervisory AI agent coordinating the entire process, and measures results by cycle time, number of human interventions, and time from approval to funds arrival. The original text cites data stating that opening an account at a Singapore private bank takes about 6 weeks, and MAS expects related initiatives to shorten the median to within one month.

    Governed case context for wealth management client onboarding, connecting CRM, KYC, screening, custody, and workflow systems with AI agents and human reviewers.

    Databricks points out that the bottleneck in wealth management client onboarding is not any single task, but rather that evidence, case status, and decisions are scattered across CRM, KYC, screening, custody, and workflow systems. Its Data + AI Platform uses governed case context to provide AI agents and human reviewers with a shared, versioned, and traceable view, with a supervisory AI agent coordinating the entire process, and measures results by cycle time, number of human interventions, and

    Matched fieldsBody

  2. OpenAI: Deployment Safety and System CardSelected

    OpenAI releases the GPT-Rosalind-6 system card addendum, offering restricted access for biological research

    OpenAI releases the GPT-Rosalind-6 system card addendum. This model uses the same set of weights as GPT-6 Astra, with no new training data or post-training added, and is provided only through trusted access to vetted research institutions, enterprises, and government partners. At the highest reasoning budget, MedChemBench v7 averages 49.6% (GPT-Rosalind-5.5: 28.9%), GeneBench Pro v13 single-attempt success rate is 37.7% (previous generation: 16.7%), and LifeSciBench Gold average score is 61.3% (previous generation: 44.0%). Under the Preparedness Framework, this release is rated High capability in the biological and chemical domains, below the Critical threshold; rated Critical in the cybersecurity domain; and below High in AI self-improvement.

    OpenAI releases the GPT-Rosalind-6 system card addendum. This model uses the same set of weights as GPT-6 Astra, with no new training data or post-training added, and is provided only through trusted access to vetted research institutions, enterprises, and government partners. At the highest reasoning budget, MedChemBench v7 averages 49.6% (GPT-Rosalind-5.5: 28.9%), GeneBench Pro v13 single-attempt success rate is 37.7% (previous generation: 16.7%), and LifeSciBench Gold average score is 61.3% (

    Matched fieldsBody

  3. Figma AI / Agent / Make

    Figma Weave in Detail: Turning Creative Processes into Reusable Workflows

    Figma Weave integrates AI generation and editing tools into reusable workflows based on a node canvas, and adds a new way of working in parallel with Figma frames. Workflows support branching exploration and comparing results from the same prompt across multiple AI models, with built-in editing nodes for color correction, compositing, levels, cropping, trimming, and painting. Weave tools are now open for community publishing, so teammates can run packaged workflows directly without manipulating nodes.

    A “Spring Arrivals” campaign design shown in Figma Design beside its connected Figma node in Weave, with options to update the Figma design or review an update.

    Figma Weave integrates AI generation and editing tools into reusable workflows based on a node canvas, and adds a new way of working in parallel with Figma frames. Workflows support branching exploration and comparing results from the same prompt across multiple AI models, with built-in editing nodes for color correction, compositing, levels, cropping, trimming, and painting. Weave tools are now open for community publishing, so teammates can run packaged workflows directly without manipulating

    Matched fieldsBody

  4. Anthropic Alignment Science Blog

    Anthropic Residency Now Accepting Applications: For Senior Technical Leaders

    Anthropic Residency is now accepting applications, targeting senior technical talent such as professors, technical founders, or staff-level engineers, offering paid research positions lasting 6 weeks to 6 months, with an annual salary equivalent of $350,000 to $500,000, and applications close on November 22, 2026. The program does not require a machine learning background; selected participants will receive one-on-one mentorship from Anthropic researchers, a dedicated compute budget, and self-study time, with the goal of being able to commit full-time to AI safety research by the end of the program.

    Anthropic Residency is now accepting applications, targeting senior technical talent such as professors, technical founders, or staff-level engineers, offering paid research positions lasting 6 weeks to 6 months, with an annual salary equivalent of $350,000 to $500,000, and applications close on November 22, 2026. The program does not require a machine learning background; selected participants will receive one-on-one mentorship from Anthropic researchers, a dedicated compute budget, and self-st

    Matched fieldsBody

  5. AWS Machine Learning

    Configure Amazon Bedrock for the ChatGPT desktop app in AWS GovCloud (US)

    The ChatGPT desktop app is now available on Amazon Bedrock in AWS GovCloud (US), accessing OpenAI frontier models through two components, Work and Codex. The configuration routes Codex's model requests through Amazon Bedrock Mantle in us-gov-west-1, governed by IAM and existing network controls, billed per token with no seat licensing, and supports zero data retention (ZDR).

    Codex CLI /status tab showing amazon-bedrock configured as the model provider

    Matched fieldsBody

  6. AWS Machine Learning

    How Worldwide Express Automated Freight Exceptions with Amazon Bedrock

    Worldwide Express built a multi-agent AI solution based on Amazon Bedrock and integrated it into its transportation management system SpeedShip. Six months after launch, active exceptions dropped by 75%, the collection cycle was shortened by 4 weeks, annual savings reached $1 million, and $10 million in free cash flow was released at once. The system dispatches invoice exceptions to three specialized agents—invoice matching, price discrepancy, and surcharge validation. Each agent first retrieves the corresponding carrier's billing rules and dispute procedures before making recommendations, and the same set of agents can be used universally without carrier-specific code.

    Exception-handling architecture: SpeedShip and other data sources feed detected exceptions to three agents on Amazon EC2 that call Amazon Bedrock, then route resolutions to a human reviewer

    Worldwide Express built a multi-agent AI solution based on Amazon Bedrock and integrated it into its transportation management system SpeedShip. Six months after launch, active exceptions dropped by 75%, the collection cycle was shortened by 4 weeks, annual savings reached $1 million, and $10 million in free cash flow was released at once. The system dispatches invoice exceptions to three specialized agents—invoice matching, price discrepancy, and surcharge validation. Each agent first retrieves

    Matched fieldsBody

Thursday

  1. OpenAI: Alignment research blogSelected

    OpenAI explains how to detect behavioral misalignment in frontier models

    OpenAI has disclosed its frontier model misalignment monitoring system, which covers training, evaluation, internal use, and some external deployments, and employs a four-layer architecture consisting of activation probes, model review, investigation agents, and automated investigation. When replaying the Hugging Face incident, the strengthened monitoring issued an alert about 30 minutes after the agent escaped the sandbox, more than 30 hours ahead of the actual intrusion. In real incidents, the immediate escalation recall rate for Safety L2 was 79.6%, and the recall rate for the Alignment L2 queue was 91.4%.

    Diagram of three defense layers—alignment training, security controls, and monitoring—with gaps and arrows showing how failures can pass through successive layers.

    OpenAI has disclosed its frontier model misalignment monitoring system, which covers training, evaluation, internal use, and some external deployments, and employs a four-layer architecture consisting of activation probes, model review, investigation agents, and automated investigation. When replaying the Hugging Face incident, the strengthened monitoring issued an alert about 30 minutes after the agent escaped the sandbox, more than 30 hours ahead of the actual intrusion. In real incidents, the

    Matched fieldsBody