Equimonet Editorial

Indian stock market research and policy explainers

Original Indian market writing built from company filings, regulator documents and attributable evidence, with sources and corrections attached.

• PhonePe gets preliminary UAE approval, with final clearance still ahead

  PhonePe says the UAE central bank has given in-principle approval for payment and stored-value licences. Commercial launch still requires final authorisation.

• Macquarie fund commits $US450 million in India’s Maple toll-road deal

  Macquarie Asset Management has completed a unit acquisition in Maple Infrastructure Trust, while both investors commit capital intended to support possible future acquisitions and expansion.

• AWS MediaLive adds frame-accurate locking without source timecode

  AWS says MediaLive can synchronize video pipelines by using visual signatures to identify and align frames, extending frame-accurate input switching to workflows without timecode from the source.

• AWS Lambda adds an explicit direct-read choice for S3 Files

  AWS says Lambda customers can now choose whether S3 Files reads come directly from an S3 bucket or use high-performance storage, separating read routing from the function’s memory setting.

• AWS adds July 2026 Supplemental Patch Bundle support to Amazon RDS for Oracle

  Amazon Web Services says Amazon RDS for Oracle supports Oracle’s July 2026 Supplemental Patch Bundle for database versions 19c and 26ai, with console-based application and a separate option for staggering automatic minor upgrades.

• AWS walkthrough pairs agent-quality scoring with infrastructure investigation

  An AWS Machine Learning Blog example shows why multi-agent systems need two monitoring layers: one to assess whether interactions meet user goals, and another to trace infrastructure failures that may appear only as degraded agent behavior.

• AWS walkthrough demonstrates MCP Apps separation between interactive UI and business logic

  An AWS Machine Learning Blog example uses Amazon Bedrock AgentCore to connect interactive HTML widgets with AI hosts, while keeping rental operations in separate Lambda and DynamoDB-backed services.

• AWS Marketplace automates demo and private‑offer requests, delivering qualified leads in minutes

  Sellers on AWS Marketplace can now receive demo and private‑offer inquiries the moment customers submit them. An automated, agentic workflow evaluates use‑case details and creates a qualified prospect in AWS Partner Central, cutting days‑long manual qualification and speeding product evaluations.

• Salesforce launches job‑ready AI agents with a long‑horizon runtime for multi‑day tasks

  Salesforce’s Agentforce portfolio now includes agents that arrive pre‑trained for specific jobs and can pursue goals over days or weeks. The new long‑horizon runtime adds memory, durable execution and dynamic steering, while early adopters report high resolution rates and rapid deployments.

• AWS adds optional unit‑test generation to Transform for .NET migration tool

  AWS Transform for .NET now includes an opt‑in capability that automatically creates unit tests for the classes it modernises. The feature is available through the AWS Toolkit for Visual Studio and works in every AWS Region where the service is offered.

• Amazon SageMaker adds prefix‑aware routing to cut LLM latency and boost cache reuse

  A new routing strategy for real‑time inference endpoints directs requests that share the same prompt prefix to the same instance, allowing key‑value cache hits to rise dramatically and reducing time‑to‑first‑token for large language models.

• Amazon Bedrock General Availability of TwelveLabs Marengo Embed 3.0 Enables Multimodal Video and Image Search

  Amazon Bedrock Knowledge Bases now offer TwelveLabs Marengo Embed 3.0 as a fully managed embedding option. The 512‑dimensional multimodal model encodes video, audio, images and text, ingesting MP4, MOV, JPEG and PNG assets directly from Amazon S3. It is generally available in US East (N. Virginia) and US West (N.

• AWS Gives API Gateway REST Logs More Size and Routing Control

  Amazon Web Services says API Gateway REST API execution logs can now reach 1 MB and be sent to multiple destinations, expanding the options for investigation, alerting and longer-term analysis.

• AWS Lambda Adds Automatic Recursive Loop Detection to Europe Sovereign Cloud

  AWS Lambda now includes built‑in protection against unintended recursive invocations for workloads running in the Europe Sovereign Cloud. The feature helps prevent runaway billing by halting loops and alerting customers through the Health Dashboard, while still allowing intentional recursion when explicitly configured.

• AWS Adds Durable Runs for Pydantic AI Agents

  The Lambda integration saves completed progress during execution, allowing Python agents to resume after interruptions without repeating finished model and tool calls.

• Amazon Quick Extends Enterprise Activity Feed to iOS and Android Devices

  Amazon has added its Quick activity feed to the mobile app for iOS and Android, giving enterprise users a single, prioritized view of work items on the go. The update preserves context across desktop and mobile, lets users act directly from the feed, and is offered at no extra charge to existing Quick customers.

• Amazon Quick adds always‑on agents, richer feed and expanded admin tools for teams

  The September update introduces cloud‑based always‑on agents that keep scheduled tasks running when a laptop is closed, a more granular activity feed with three daily briefings and seven‑day search, and new enterprise controls such as mobile device management and Microsoft Purview DLP integration.

• AWS Launches General Availability of Second‑Generation Single‑Rack Outposts

  Amazon Web Services announced the GA of a 42‑U single‑rack Outpost that packs up to 2,688 vCPUs and 100 TB of Amazon EBS storage. The compact unit runs the same AWS APIs, console and security controls as multi‑rack Outposts and is aimed at locations with limited rack space and power.

• Amazon MQ adds RabbitMQ 4.3 with expanded quorum‑queue features

  Amazon MQ now supports RabbitMQ version 4.3 on m7g brokers in every region where m7g instances are offered. The update brings disk‑space compaction, 32 strict priority levels, native delayed retries and per‑protocol consumer timeouts for quorum queues, while retiring several legacy queue types.

• Amazon CloudWatch adds Network Health Indicator for Transit Gateway inter‑Region peering

  Synthetic monitors in Amazon CloudWatch now emit a Network Health Indicator for traffic that traverses a Transit Gateway inter‑Region peering connection, helping operators tell whether latency originates in the AWS network. The capability is live in all commercial AWS Regions except GovCloud (US) and China.

• Android adds built‑in password‑manager migration for secure, file‑free credential transfers

  Google’s Android platform now includes a native experience that moves passwords and passkeys between supported managers without creating unencrypted files. The workflow coordinates the hand‑off, keeps data encrypted, and lets users switch tools in a few taps.

• Google Releases Gemini AI Desktop App for Windows 10 and 11

  Google has released the Gemini AI assistant as a native Windows application for Windows 10 and 11. The app launches with an Alt + Space shortcut and offers document drafting, task management, and on‑demand image or video generation directly from the desktop.

• AWS example shows how prompt-based PII detection can adapt across models and domains

  An AWS Machine Learning Blog walkthrough evaluates a configurable large-language-model detector on five public corpora, highlighting how changing the entity schema can target domain-specific PII without retraining.

• Turn‑Level Correctness Metric Isolates Errors in Multi‑Turn AI Agents

  The Agent Evaluation Metric (AEM) breaks down correctness into truthfulness and completeness for every turn of a conversation, letting developers pinpoint the exact step where a mistake first appears and see how it propagates through later turns.

• AWS Elemental Inference Adds Real‑Time Contextual Metadata to Live Video Streams

  AWS now lets broadcasters and content platforms generate scene‑level metadata as live video is encoded. The serverless AI extracts taxonomy categories, brand‑suitability signals, objects, actions and shot descriptions, feeding directly into MediaLive and MediaTailor for smarter ad targeting and automated asset enrichment.

• AWS Elemental MediaLive Introduces A/B Forensic Watermarking for Live Video Protection

  MediaLive can now generate two synchronized live streams, each carrying a distinct, visually transparent watermark that survives re‑encoding and screen capture.

• AWS MediaTailor Introduces Free Yield Optimization to Monetize Unfilled Live‑Stream Ads

  AWS Elemental MediaTailor now offers a Yield Optimization capability that automatically inserts Amazon Ads into ad breaks that would otherwise stay empty during server‑side ad insertion for live events.

• GeForce NOW’s New Releases Show Why Launch Timing Matters

  NVIDIA’s latest GeForce NOW lineup illustrates the difference between a game’s store release, its early-access status and its availability on the cloud gaming service.

• Google adds integrations and measurement tools to its advertising stack

  Google is connecting Data Manager with Analytics and Display & Video 360 and expanding Meridian. Its reported advertising uplifts are company-observed averages, not promised results for every advertiser.

• Salesforce adds Fin’s AI customer‑agent platform to boost autonomous service

  Salesforce says it completed its acquisition of Fin, formerly Intercom, on September 10, 2026. It reports an average 76% resolution rate for Fin’s AI customer agent; this does not mean every customer achieves that rate.

• IBM and NASA Release Open-Source Lunar AI Model with Notable Accuracy Gains

  IBM and NASA have made the NASA‑IBM Lunar Foundation Model publicly available, pairing it with a unified, machine‑learning‑ready lunar dataset. The model shows measurable improvements in detecting ice deposits, volcanic features and craters, offering researchers a shared AI foundation for Moon science.

• Southbank Centre to host four Google Arts & Culture experiments on AI and creativity

  Google Arts & Culture and London’s Southbank Centre will present four interactive experiments during the inaugural Creative Intelligence weekend, examining how artists and digital tools can shape creative experiences while keeping human agency in view.

• Amazon Expands Kiro Access Across 132 Universities

  Amazon says students at 132 universities across 18 countries can receive a year of free Kiro access, including 1,000 credits per month and tools designed around structured software development.

• AWS gives identity-matching workflows a confidence score for every record

  AWS says Entity Resolution now assigns individual confidence scores to records matched through machine-learning workflows, allowing users to apply different thresholds to automated merges and downstream activation.

• NVIDIA teams with Australian data‑center partners to target up to 2 GW of AI factory capacity by 2027

  NVIDIA disclosed a collaboration with eight Australian cloud and infrastructure partners to expand land, power and shell capacity for its DSX AI factories. The plan aims for a build‑out of up to 2 GW by 2027, providing high‑performance compute for startups, universities, researchers and enterprises across the region.

• Amazon EBS Volume Clones Add Cross‑Account Copy with Re‑Encryption

  Amazon Elastic Block Store now lets customers copy EBS volumes between AWS accounts, optionally re‑encrypting them with a KMS key in the target account. The feature works for all volume types and is available in every region that supports Volume Clones.

• AWS Walkthrough Shows How Qwen3.8-2.4T-A95B Fits a Single 8-GPU Node

  The dated technical example focuses on the infrastructure and serving choices needed to run Alibaba’s open-weight model, including NVFP4 quantization, vLLM configuration and HyperPod capacity planning.

• AWS Systems Manager adds six diagnostic categories to reduce unmanaged EC2 instances

  AWS Systems Manager now expands its diagnosis capability to pinpoint six additional reasons why Amazon EC2 instances and hybrid‑activated nodes remain unmanaged, helping customers bring fleets under control faster with actionable guidance and built‑in automation.

• AWS Lambda adds Graviton5 support to Managed EC2 instances, boosting performance and cost efficiency

  AWS Lambda now lets customers run functions on the latest Graviton5‑powered C9g, C9gd, M9g and M9gd EC2 instances through Lambda Managed Instances. The new option promises up to 25% higher compute performance versus Graviton4 while preserving Lambda’s fully managed operational model, offering a clear path to faster, cheaper serverless workloads.

• Heurist Finance cuts engineering effort by 80% with Amazon Bedrock AgentCore payments

  Heurist Finance uses Amazon Bedrock AgentCore to orchestrate AI‑driven research while buying premium market data only when needed. The managed infrastructure handles identity, memory, sandboxed code execution and per‑query crypto payments, letting the startup focus on retail‑grade portfolio analysis.

• Amazon Bedrock adds APIs and console tools to debug document‑level access control

  Amazon Bedrock Managed Knowledge Base now includes the CheckIngestedDocumentAcl and GetIngestedDocumentAcl APIs, together with a new Document Access Control console section. Administrators can verify individual user permissions on ingested documents and audit full ACLs without opening a support case.

• Amazon expands Nepal flood relief with cash grants and emergency supplies

  In response to flash floods and landslides that have impacted more than 93,000 people in Nepal, Amazon is providing additional cash donations to seven nonprofit partners and preparing over 43,000 emergency items such as tents, blankets and solar lights for free delivery to first‑responders.

• Apple Watch Ultra 4 Adds a Daily Readiness Score for Training Decisions

  Apple says its new readiness feature combines recent activity, training load, vitals and sleep score into a changing 0-10 assessment, while separate sensor updates increase measurement frequency.

• Apple Watch Series 12 Turns Higher-Frequency Health Data Into a Daily Readiness Score

  Apple says its new watch combines more frequent heart-rate and HRV measurements with activity, training load and sleep data to produce a 0-10 readiness score, while several related Health app features are planned for later this year.

• iPhone 18 Pro Puts Camera Control and Image Authenticity at the Center

  Apple says its new Pro models combine a variable-aperture 48MP Fusion Main camera with Apple Reference Image, giving users more control over capture and a way to compare an image with signed sensor data.

• AWS Builder ID adds recovery paths for third-party sign-ins

  AWS says Builder ID users can add a recovery email, regain access through self-service options and register MFA devices when signing in with Google, Apple, GitHub or Amazon.

• How Samsung Says It Cut Galaxy Z Fold8’s Weight to 201 Grams

  Samsung attributes the Galaxy Z Fold8’s lower weight to cumulative changes across its display, hinge, materials and hundreds of internal parts—not to a single redesigned component.

• Amazon Timestream for InfluxDB 3 adds managed custom Python plugins

  AWS says users can run workload-specific Python plugins on managed InfluxDB 3 Core and Enterprise editions, with repository, parameter-group, trigger and authentication steps kept distinct.

• NVIDIA Pushes Synthetic-Video Checks Into Broadcast and Compliance Workflows

  NVIDIA says its Synthetic Video Detector has reached 99.3% accuracy for text-to-video content and 97.7% for image-to-video content, while partners are embedding its signals into review and streaming tools.

• How human direction shaped an AI recreation of a 70-year love story

  Google DeepMind says the short documentary Love, Rendered combines restored photographs with performance-capture models to recreate an unrecorded meeting recalled by a couple married for over 70 years.

• How Salesforce Scaled Agentic AI Coding Across 15,000 Engineers

  Salesforce moved from a 30‑day pilot with 200 engineers to an enterprise‑wide rollout that now touches 15,000 developers.

• Prime Video launches AI‑powered lip‑sync dubbing on Maxton Hall to align mouths with dialogue

  Prime Video introduced a new lip‑sync feature that aligns human‑dubbed audio with actors' mouth movements. The technology debuts on the International Original series Maxton Hall, is live for Seasons 1‑2, and will be included when Season 3 launches on Dec 9, with plans to expand to other titles.

• Target Unveils ‘Target by Design’ Immersive Event Showcasing Accessible Design

  Target Corporation is hosting a free, two‑day immersive presentation called Target by Design on Sept. 10‑11 at the New Museum’s Bowery space. The experience reimagines everyday items such as carts and bags, highlights historic collaborations, and offers emerging creators a platform to showcase their work.

• CFOs Turn to AI Agents to Tackle Growing Revenue‑Model Complexity

  A new Salesforce CFO Priorities Report shows that finance leaders are expanding their roles to govern AI while using it to automate fragmented revenue processes across multiple channels and subscription models. While 90% of AI users report a positive return, security and governance remain the chief obstacles to broader adoption.

• Google commits €13 billion to AI and clean‑energy infrastructure in Finland

  Google is investing €13 billion over the next two years to expand AI‑focused data centres, boost clean‑energy capacity and fund community programmes across Finland. The plan promises thousands of permanent jobs, a short‑term construction boost of tens of thousands of workers and new training pathways for Finnish talent.

• Google’s Finnish Energy Partnerships Aim to Power Data Centers with Clean, Affordable Electricity

  Google announced a suite of energy agreements in Finland, including a nuclear plant life‑extension, new wind contracts and a 94 MW battery, to support its expanding data‑center footprint while keeping electricity prices stable for households and businesses.

• How GPT‑6 Astra on Amazon Bedrock Enables Secure, High‑Performance AI for Complex Business Tasks

  Amazon Bedrock now offers the generally available GPT‑6 Astra model from OpenAI. The model adds deeper reasoning, explicit prompt caching and enterprise‑grade security, letting organizations run AI agents for finance, contract review and software development at production scale while keeping data private and auditable.

• IBM, Cleveland Clinic and RIKEN Reach Quantum‑Centric Supercomputing Milestone in Protein Simulation

  A joint team from Cleveland Clinic, Japan’s RIKEN Center for Computational Science and IBM has been named a finalist for the 2026 ACM Gordon Bell Prize after simulating a 12,635‑atom protein using a hybrid quantum‑classical workflow that blends IBM’s Heron quantum processors with the Fugaku and Miyabi‑G supercomputers.

• AWS adds event‑based retention to S3 Object Lock for flexible WORM protection

  Amazon Web Services announced that S3 Object Lock can now use event holds, letting customers start a write‑once‑read‑many retention period after a specific trigger such as a contract signing or audit completion. The feature is available in all regions at no extra cost and includes new IAM and bucket‑policy controls.

• AWS releases Nx Plugin for AWS to scaffold secure full‑stack apps

  AWS announced version 1.0 of the open‑source Nx Plugin for AWS, a toolkit that extends the Nx monorepo build system with generators for AI agents, APIs, websites and databases.

• SageMaker Feature Store adds feature‑level writes to cut latency and cost

  Amazon SageMaker Feature Store now lets data scientists update individual feature values in a record with a single request, avoiding full‑record rewrites. The new feature‑level write capability reduces write latency and expense and is available in every AWS Region where the service is offered.

• OpenAI’s GPT‑6 Astra Now Generally Available on Amazon Bedrock, Offering Million‑Token Context

  Amazon Web Services announced that OpenAI’s GPT‑6 Astra model is generally available on Amazon Bedrock. The model supports up to one million input tokens, can be invoked through Bedrock APIs or the console, and includes new enterprise plugins that extend browser‑use capabilities for business applications.

• Samsung teams up with Mistral AI to embed on‑premise large language models in semiconductor manufacturing

  Samsung Electronics announced a strategic partnership with Mistral AI to integrate the startup’s Large language model into its semiconductor design and fab operations.

• AWS Transform Expands to GovCloud US‑West, Streamlining Government Server Migrations

  AWS has added its Transform migration service to the AWS GovCloud (US‑West) region, giving federal agencies and regulated firms a way to automate large‑scale server moves within a secure, isolated environment.

• Amazon Bedrock AgentCore Memory adds IngestData API for direct long‑term storage

  Developers can now bypass short‑term events when adding information to AgentCore's long‑term memory using the new IngestData API, which accepts conversational or JSON payloads and is available in every AWS Region where AgentCore Memory is supported.

• Amazon SageMaker Feature Store adds UpdateRecord for efficient feature‑level writes

  The new UpdateRecord API lets developers modify one or more feature values in a SageMaker Feature Store record without the costly read‑modify‑write cycle. Available on both DynamoDB‑backed Standard and ElastiCache‑backed In‑Memory online stores, it reduces latency, cuts read‑capacity charges, and adds fine‑grained IAM controls.

• AWS adds more precise cropping controls and a testing playground to CloudFront image transformations

  AWS says Dynamic Image Transformation for Amazon CloudFront now supports combined detection methods, configurable cropping controls, broader automatic optimization, an interactive validation playground and feature parity across its ECS and Lambda architectures.

• Slackbot’s Dreamforce Advice Puts Meeting Priority Ahead of a Packed Calendar

  Salesforce’s Ask Slackbot column draws on an account executive’s Slack conversations and related documents to offer a focused approach to Dreamforce scheduling: rank important accounts, protect travel time and confirm key appointments.

• AWS walkthrough example shows hybrid cross-account model governance

  An AWS technical walkthrough demonstrates how regulated environments can isolate development accounts while moving only approved, self-contained model copies into a central governance account.

• A SageMaker AI walkthrough shows why LLM inference choices depend on the workload

  The source’s example compares GPU configurations for two 30B Mixture-of-Experts models, showing how throughput, latency and estimated cost can point to different deployment choices.

• AWS walkthrough shows how an agent CI gate can catch regressions before merge

  An AWS technical example combines AgentCore, MCP authorization and GitHub Actions to score agent behavior in a development environment and block pull requests when results fall below a threshold.

• Walkthrough: How HPE Zerto designed an on-premises AI troubleshooting system

  An AWS and HPE Zerto technical example shows how local agents, live operational data and Amazon Bedrock can be combined for controlled infrastructure troubleshooting.

• AWS Walkthrough Shows How DiDi’s Example QA System Uses Context Management

  An AWS technical post describes DiDi’s example of a multilingual contact-center quality-assurance system, where separating tasks and controlling model context supported reported gains in intent verification and compliance scoring.

• Missouri’s Google AI partnership puts adoption decisions with local schools

  Google says the statewide initiative will give nearly 100,000 educators and more than 1.1 million students no-cost access to AI tools, training and credentials, while participation remains voluntary for institutions.

• Samsung and ASML Expand High NA EUV Collaboration Around 12-Inch Masks

  Samsung will join work on next-generation 12-inch photomask technology and plans to introduce ASML’s High NA EUV in future DRAM high-volume manufacturing by 2028.

• Zero‑Coupon Bonds: Deep‑Discount Debt, Tax Implications, and Market Behavior

  Zero‑coupon bonds trade far below face value, defer all interest until maturity, and can be issued by governments or corporations. Their prices are more volatile and holders must report phantom interest each year.

• Google’s Route 66 AI tour points small businesses toward hands-on training

  Grow with Google says its August workshops reached small business owners across five states, while free online training remains available to U.S. small businesses after the tour ended.

• Understanding State Securities Regulators and Their Role in U.S. Investment Oversight

  State securities regulators enforce “blue‑sky” laws that complement federal securities regulation. They oversee local securities offerings, register smaller investment advisers, and provide resources for investors to verify broker and adviser histories.

• Securities Lending Explained: How Institutions Earn Fees by Temporarily Sharing Assets

  Securities lending is a market practice where large institutional investors temporarily transfer securities to brokers or dealers in exchange for a fee and cash collateral. The lender can earn fee income and may invest the collateral, while borrowers use the securities for market‑making, short sales, or to meet delivery obligations.

• How Reverse Stock Splits Work and Why Companies Use Them

  A reverse stock split consolidates a company’s existing shares into fewer, higher‑priced shares. Companies may employ it to lift a low trading price or satisfy exchange listing rules, while state law and corporate bylaws dictate the approval process and SEC filings disclose the action.

• How Restricted Securities Work and When They Can Be Sold Publicly

  Restricted securities are privately placed, unregistered securities that carry a legend preventing public resale. Learn how the legend is removed, the role of transfer agents, and the key requirements of Rule 144 for eventual public sale.

• How to Verify Who Really Owns Your Securities

  Understanding the steps to prove ownership helps you locate records, navigate broker and transfer‑agent responsibilities, and avoid losing assets to state escheatment. This guide outlines the definition, where records are kept, and the key distinctions you need to know.

• Mutual Fund Prospectus: Required Content and How to Access It

  Mutual‑fund prospectuses must be provided after a share purchase, yet investors can request them beforehand. The SEC mandates two formats—a detailed statutory prospectus and a concise summary—each presented in a standard order to help investors compare funds.

• Lost or Stolen Securities Certificates: Steps for Reporting and Replacement

  Investor.gov describes immediate notification, stop-transfer requests, replacement requirements and the distinction between physical certificates and book-entry or street-name holdings.

• Financial Planners: Understanding Scope, Services and Compensation

  Financial planners may provide broad financial planning or narrower product recommendations. Understanding their services, payment arrangements and professional backgrounds helps clarify what the role includes.

• Escheatment by Financial Institutions: What Happens to Unclaimed Assets

  Escheatment is the process through which a state takes custodial responsibility for financial assets that remain unclaimed after efforts to locate the owner have been unsuccessful.

• Convertible Securities: How Conversion Formulas Shape Dilution

  Convertible securities combine a bond or preferred stock with a possible conversion into common shares. The conversion formula—fixed or linked to market prices—helps determine how future share issuance may affect ownership and value.

• Closed-End Funds: How Their Structure Shapes Shares, Portfolios and Trading

  An evergreen guide to what closed-end funds are, how the source describes their regulation and management, and why share trading status differs from portfolio liquidity.

• What Happens When a Bond Is Sold Before Maturity

  Selling a bond before it comes due can produce a different amount from its face value. The outcome depends on market interest rates and transaction charges described by Investor.gov.

• How fund distribution expenses differ from shareholder-service costs

  This background note distinguishes the costs of marketing fund shares from the costs of answering shareholder questions, using an Investor.gov educational reference from the United States.

• Why Funds That Invest in Other Funds Can Have Layered Expenses

  A fund of funds bears the operating costs of its own structure while indirectly bearing expenses charged by the funds it owns. The distinction explains why one investment can involve more than one layer of fund expenses.

• Understanding the Difference Between Brokerage Commissions and ETF Expense Ratios

  ETF expense ratios are disclosed as a percentage of a fund’s average net assets and are paid out of the fund’s assets, lowering the net asset value. Brokerage commissions are flat fees paid to a broker each time ETF shares are bought or sold and do not appear in the fund’s prospectus fee table.

• How Portfolio Income Differs From Capital‑Gains Distributions in Mutual Funds

  Mutual funds provide investors with two distinct types of payouts. Portfolio income comes from dividends and interest earned on the fund’s holdings, while capital‑gains distributions arise when the fund sells securities at a profit and passes the net gain to shareholders.

• A zero expense ratio does not necessarily mean zero investing costs

  An evergreen explanation of the difference between a fund’s stated expense ratio and other costs that may accompany investing in it.

• Understanding Recurring Fund Operating Expenses vs. Shareholder Fees

  Mutual funds and ETFs disclose two broad categories of costs in their prospectus fee tables: recurring operating expenses that are paid out of fund assets, and shareholder fees that are charged directly to investors when they buy, sell, or maintain their accounts.

• Samsung Rolls Out AI‑Powered Software Updates to Extend Appliance Lifespan

  Samsung Electronics is launching Tizen OS 10.0 updates for selected refrigerators and laundry appliances. The upgrades add Google Gemini‑based AI Vision for food management and a refreshed Bixby voice experience, aiming to keep devices functional and valuable throughout their ownership period.

• Amazon SageMaker Batch Transform adds GPU‑rich G6e instance support

  SageMaker’s offline inference service now runs on Amazon EC2 G6e instances, which pack up to eight NVIDIA L40S Tensor Core GPUs and third‑generation AMD EPYC CPUs. The update targets large language and diffusion models and is available in several AWS regions.

• Intuit’s EWOK Agent: AI‑Powered Disaster Recovery Built on Amazon Bedrock

  Intuit extended its internal disaster‑recovery orchestrator, EWOK, with an AI plug‑in that leverages Amazon Bedrock foundation models.

• How Amazon Bedrock and Textract Enable Scalable Knowledge Bases for Complex Utility Bills

  Customer‑service teams that receive thousands of utility statements each month can now automate extraction and querying of those documents.

• InstantStart adds an AI‑driven control plane for Amazon SageMaker HyperPod cluster provisioning

  Amazon SageMaker HyperPod’s InstantStart project packages the multi‑step bootstrap of HyperPod clusters into a single out‑of‑band container. It offers a web UI, a REST API and an AI‑agent interface that all invoke the same backend, while keeping every created AWS and Kubernetes resource visible and inspectable.

• C3 AI generated cash and still made a loss. Both numbers are real

  The enterprise-AI company's quarterly results offer a useful lesson in reading earnings: a positive cash-flow headline does not replace the income statement.

• NVIDIA agreed to buy Hugging Face for $12.93 billion. The open-platform promise is now the story

  The agreement joins the largest AI-compute supplier with a platform used to share more than three million models. NVIDIA says Hugging Face will remain open and multi-cloud; the SEC filing shows why that promise, staff retention and regulatory clearance matter before the deal closes.

• Hindustan Zinc's rare-earth LoI is real—but it is not a producing mine

  Karnataka has issued a letter of intent for a 314.21-hectare REE and yttrium block. The milestone is genuine, but the one-page filing does not establish reserves, capex, production or earnings. Here is the gap between allocation and cash flow.

• NSE IPO is a 100% offer for sale: what the draft documents actually show

  NSE has filed draft offer documents for a public listing, but this is not a fresh-capital raise. Here is the verified offer structure, the shareholders selling, and every major term that is still blank.

• What SEBI’s FY26 derivatives study says about intensity, persistence and loss

  SEBI’s latest evidence does more than count losing traders: it maps how options concentration, capital employed, trading frequency and repeated participation relate to outcomes.

• The signal was not the rate: how RBI framed the August pause

  The repo rate stayed at 5.25%, but the more useful reading lies in the MPC’s inflation composition, conditional forecasts and preference for flexibility.

• What the RBI’s ₹1.5 lakh crore VRRR does—and does not—say

  A variable-rate reverse-repo auction is an operating instrument for absorbing liquidity. Reading it as a standalone market forecast goes beyond the document.

• The denominator change inside the latest priority-sector lending amendment

  RBI’s amendment changes how specified advances backed by eligible long-tenor non-resident deposits enter adjusted net bank credit for priority-sector targets.