Today’s tech highlights include Google’s launch of Gemini 2.5, surpassing benchmarks in UI interaction, and the reveal of Gemini Enterprise for workplace automation. Meanwhile, the AI market faces a potential correction amid oversupply, and Microsoft closes a Windows 11 loophole for account-free installs. Major infrastructure investments, such as OpenAI’s $100 billion data center deal, underscore AI’s macroeconomic impact.
▶️ Internet Infrastructure
Google DeepMind Launches Gemini 2.5 for Advanced UI Interaction
Google DeepMind’s Gemini 2.5 Computer Use model powers UI-interacting agents, surpassing benchmarks in web/mobile control with lower latency, accessible via Gemini API on Google AI Studio and Vertex AI.
- Google DeepMind released the Gemini 2.5 Computer Use model via the Gemini API on October 7, 2025
- Built on Gemini 2.5 Pro, it enables agents to interact with user interfaces, outperforming competitors in web and mobile control benchmarks with lower latency
- Available in preview on Google AI Studio and Vertex AI, supporting tasks like form filling, clicking, typing, scrolling, and login operations
Fly.io CEO Phished in Sophisticated Attack Highlights MFA Importance
Fly.io CEO Kurt Mackey was phished through a sophisticated email, causing a Twitter account takeover; the incident highlights the importance of phishing-resistant MFA and account hygiene.
- Fly.io CEO Kurt Mackey was phished via a convincing email exploiting his middle-aged insecurity, leading to account compromise.
- The attack resulted in Twitter account takeover (ATO), with attacker revoking tokens and setting up new 2FA, taking approximately 15 hours for recovery.
- Fly.io secures infrastructure with phishing-resistant MFA (FIDO2, Passkeys) via Google IdP; Twitter account was a legacy shared account managed in 1Password.
Data Center and AI Investment Drive 92% of Early 2025 U.S. GDP Growth
U.S. GDP growth in early 2025 was driven mainly by data center and AI infrastructure investment, which accounted for 92% of growth, highlighting technology’s pivotal macroeconomic role.
- U.S. GDP growth in first half of 2025 was only 0.1% excluding data center and information processing tech investment, which drove nearly 92% of GDP growth
- Investment in data centers and AI infrastructure surged, with hyperscalers spending nearly $400 billion annually, contributing approximately 100 basis points to U.S. GDP growth
- Data-center buildout and AI demand caused a surge in infrastructure spending, with tech giants like Microsoft, Google, Amazon, Meta, and Nvidia investing tens of billions
Bluesky Federation Faces Challenges in Decentralization and Moderation
Bluesky’s federation via ATProto remains incomplete, with Blacksky dependent on Bluesky’s infrastructure; recent moderation issues highlight ongoing decentralization and moderation challenges.
- Bluesky’s federation system relies on a complex ATProto protocol, with key components including Personal Data Servers (PDS), Relay servers, Labelers, and AppViews
- Blacksky, an independent ATProto implementation, remains dependent on Bluesky’s application server and moderation tools due to software limitations
- A recent ban of a Blacksky user, Link, by Bluesky moderators exposed the platform’s incomplete decentralization and moderation transparency issues
Roman Zipp Advocates Self-Hosting to Boost Privacy and Sovereignty
Roman Zipp promotes self-hosting personal services like CalDAV, CardDAV, email, and smart home systems to improve privacy and sovereignty, detailing technical options and security considerations.
- Roman Zipp advocates for increased self-hosting of personal services to enhance privacy and sovereignty, discussing applications like CalDAV, CardDAV, mail, smart home, RSS, and location tracking.
- Emphasizes privacy benefits by hosting own calendar, contacts, and email servers, and highlights tools such as Radicale, Baïkal, Mailcow, and Home Assistant.
- Addresses security concerns, including DDoS risks, and suggests solutions like Cloudflare tunnels; discusses challenges of self-hosted email deliverability and privacy implications of big tech.
RubyCentral AWS Root Account Breached for 11 Days, Security Enhanced
Ruby Central’s AWS root account was accessed by an unauthorized actor from September 19-30, 2025, leading to credential changes; control was restored with no data compromise, and security protocols were strengthened.
- Unauthorized actor accessed RubyGems.org AWS root account from September 19 to September 30, 2025, changing the password and removing authorized users.
- Ruby Central regained control on September 30 by resetting the root password, revoking credentials, and implementing enhanced security measures.
- No evidence indicates user data, gems, or infrastructure were compromised; the incident was treated as a security breach due to unrotated credentials and continued access disclosure.
Low-Latency 10G Ethernet FPGA Core Achieves Sub-60ns Loopback
Tom Chisholm’s series details developing a sub-60ns latency 10G Ethernet FPGA core, emphasizing verification, latency techniques, and performance analysis.
- Series of blog posts on developing a low latency 10G Ethernet FPGA core, starting with introduction on May 1, 2023
- Achieved less than 60ns loopback latency, comparable to commercial solutions
- Focuses on verification with cocotb and pyuvm, latency reduction techniques, and analysis of low latency Ethernet cores
Firefly to Acquire SciTec to Boost Defense Tech and Missile Defense Capabilities
Firefly Aerospace intends to acquire SciTec for $855 million to strengthen its defense tech offerings, supporting projects like the $3.6 trillion Golden Dome missile interception system.
- Firefly Aerospace plans to acquire defense analytics firm SciTec for $855 million in cash and equity, with closing expected by the end of 2025
- The acquisition aims to enhance Firefly’s capabilities in defense missions like Trump’s Golden Dome missile interception system, projected to cost up to $3.6 trillion over 20 years
- SciTec specializes in missile defense, remote sensing, and command and control systems, with a $259 million US Space Force contract awarded in May and annual revenue of approximately $164 million
Serve Robotics Teams Up with DoorDash to Expand Robot Delivery Nationwide
Serve Robotics partners with DoorDash to deploy delivery robots in Los Angeles, asserting robot delivery surpasses robotaxis in market size and aims for a shared infrastructure model.
- Serve Robotics will begin delivering for DoorDash in Los Angeles, expanding to nationwide rollout.
- Serve operates delivery robots in five cities: Atlanta, Chicago, Dallas, Los Angeles, and Miami; aims to collaborate with multiple delivery services and retailers.
- CEO Ali Kashani states delivery by robots is a larger business opportunity than robotaxis, emphasizing the potential for a shared platform infrastructure.
GitHub Fixes Webhooks Performance Issue After Network Glitch
GitHub resolved a Webhooks degraded performance incident affecting multiple services, with full recovery by 16:40 UTC on October 9, 2025, after identifying and removing a faulty network component.
- GitHub resolved a Webhooks performance degradation incident reported on October 9, 2025, at 14:45 UTC.
- The incident affected GitHub Services including Git Operations, Webhooks, API Requests, Issues, Pull Requests, Actions, and Pages.
- Full recovery was achieved by 16:40 UTC, with all systems operating normally and delays resolved.
Synology Reverses Policy, Restores Support for Third-Party HDDs and SSDs
Synology reinstated support for third-party HDDs and SSDs in DSM 7.3 after NAS sales declined due to earlier restrictions, reversing a policy that limited drive compatibility and damaged reputation.
- Synology reversed its policy banning third-party HDDs and 2.5-inch SATA SSDs with the release of DSM 7.3
- The policy, introduced earlier in 2025, restricted drives from brands like Seagate and WD in models such as DS925+, DS1825+, and DS425+
- The change allows third-party drives to function fully with monitoring, alerts, and storage features, restoring user choice and reducing costs
MAME Completes Hyper Neo Geo 64 Sound Support After 21 Years
MAME finally added sound emulation for Hyper Neo Geo 64, completing support after two decades, with recent updates significantly improving audio fidelity and preparing for release in version 0.282.
- MAME developers achieved sound support for Hyper Neo Geo 64 after 21 years of development, with recent updates in MAME 0.281 and 0.282.
- The system was active from 1997 to 1999, with seven games: Road’s Edge, Samurai Shodown 64, Xtreme Rally, Beat Busters: Second Nightmare, Samurai Shodown 64: Warriors Rage, Fatal Fury: Wild Ambition, Buriki One.
- Recent improvements include fixing volume envelopes, sample looping, and sound filtering, culminating in fully playable audio in the upcoming MAME 0.282 release.
Cisco Unveils 8223 Router with 3 Exabits/sec for AI Data Centers
Cisco’s 8223 router, powered by Silicon One P200 ASIC, supports up to 3 exabits/sec, linking datacenters up to 1,000 km apart for AI training at scale.
- Cisco announced the 8223 router, a 51.2 Tbps device powered by Silicon One P200 ASIC with 800 Gbps coherent optics supporting up to 1,000 km span
- The architecture can theoretically achieve 3 exabits per second aggregate bandwidth by linking multiple routers, enabling multi-site AI training clusters with several million GPUs
- Smaller configurations support up to 13 Pbps bandwidth using a two-tiered network; large deployments require thousands of routers, with cloud providers like Microsoft and Alibaba evaluating the chips
IBM Teams Up with Cockroach Labs to Bring Cloud-Native PostgreSQL to Mainframes
IBM partners with Cockroach Labs to offer CockroachDB on mainframes and hybrid cloud, enabling cloud-native PostgreSQL deployment for legacy systems, with concerns over migration complexity and Db2’s continued dominance.
- IBM signed an OEM agreement with Cockroach Labs to integrate CockroachDB, a distributed PostgreSQL-compatible cloud RDBMS, into mainframe environments.
- The partnership targets hybrid cloud deployments across IBM LinuxONE, Linux on Z, Power Systems, and Red Hat OpenShift, enabling migration without rewriting core applications.
- Experts note that while CockroachDB offers cloud-native resilience, IBM’s mainframe Db2 remains the primary choice for large enterprise workloads, and migration risks may deter existing Db2 users.
China Tightens Export Controls on Rare Earth Technologies and Goods
China imposed laws requiring licenses for exporting rare earth technologies and goods containing Chinese-sourced rare earths, expanding control over global supply chains and mimicking US export restrictions.
- China’s Ministry of Commerce announced laws requiring licenses for exporting rare earth-related technologies and goods containing Chinese-sourced rare earths outside China.
- New regulations extend export restrictions to items manufactured abroad that contain rare earths sourced from China, with licensing required for dual-use and advanced semiconductor-related items.
- The measures mimic US export controls, including case-by-case approval for technologies related to rare earth mining, processing, and manufacturing, especially for military and AI applications.
Google Launches Gemini Enterprise for AI-Driven Workflow Automation
Google launched Gemini Enterprise, a rebranded platform based on Agentspace, enabling AI-driven workflow automation with no-code tools, integrations, and CLI extensions to enhance enterprise productivity.
- Google announced Gemini Enterprise on October 9, 2025, as a platform for automating business workflows using Google’s Gemini ML models.
- Rebrands and extends Google Agentspace, now accessible via Gemini Enterprise page, with a web interface, no-code workbench, and integrations.
- Features include AI-powered conversational interface, pre-built and customizable software agents, connections to Google Workspace and Microsoft 365, and Gemini CLI extensions for external tool integration.
Intel Panther Lake 18A Chip to Feature 8-16 Cores and 180 TOPS AI
Intel’s Panther Lake, based on its 18A process, features multi-die architecture with 8-16 cores, up to 180 TOPS AI, and supports 128 GB DDR5, with full details pending before January 2026 release.
- Intel began volume production of Panther Lake’s compute tiles on its 18A process at Fab 52 in Arizona during summer 2025
- The notebook chip, launching in January 2026, offers 8 to 16 cores, up to 180 TOPS AI performance, and supports up to 128 GB DDR5 at 7,200 MT/s
- The processor architecture is multi-die, combining a compute tile on 18A, a TSMC platform controller, and a GPU tile on either Intel 3 or TSMC, using Foveros packaging
Microsoft Blames Kubernetes Crash for October 2025 Azure Outage
Microsoft blamed a Kubernetes crash, not a deployment error, for the October 9, 2025, Azure Front Door outage affecting 30% of instances, with services gradually restored through restarts.
- Microsoft attributed the outage to a Kubernetes crash, not a faulty deployment, causing approximately 30% capacity loss of Azure Front Door instances globally, mainly in Europe, Middle East, and Africa.
- The incident began at 0740 UTC, with Microsoft restarting underlying Kubernetes instances to restore services; most impacted resources have been recovered.
- Microsoft did not specify the root cause of the Kubernetes crash or why automatic recovery failed, but confirmed that the majority of affected services, including Azure Portal and Microsoft Entra, are now operational.
SoftBank Acquires ABB Robotics to Boost Industrial AI and Automation
SoftBank’s $5.4 billion acquisition of ABB Robotics aims to advance “physical AI” by integrating industrial robotics with artificial superintelligence, supporting its broader automation and AI strategy.
- SoftBank Group acquired ABB’s robotics division for $5.4 billion, announced on October 8, 2025, pending regulatory approval
- ABB will establish a new holding company for its robotics unit before transferring ownership to SoftBank
- ABB Robotics employs approximately 7,000 people, generated $2.3 billion in revenue last year, and specializes in industrial robotic arms for assembly, painting, and packaging
SonicWall Cloud Backup Breach Exposes All User Data
SonicWall’s cloud backup service was compromised, exposing configuration files for all users, prompting urgent security measures and infrastructure hardening.
- SonicWall confirmed all users of its MySonicWall cloud backup service were affected by a cybersecurity incident, not just the initially claimed 5%
- Attackers accessed and potentially exfiltrated encrypted, compressed firewall configuration backup files stored on SonicWall’s systems
- The breach impacted every customer who used the cloud backup service, regardless of backup creation date, following an investigation and external forensic review
▶️ Open Source
Meta Launches React Foundation for Community-Led Innovation and Support
Meta announced the formation of the React Foundation within the Linux Foundation to oversee React and React Native, ensuring independent governance, community support, and continued innovation with Meta’s $3 million funding and engineering support.
- Meta open-sourced React over a decade ago; now transitioning React, React Native, and supporting projects like JSX to the React Foundation.
- The React Foundation, part of the Linux Foundation, will maintain infrastructure, organize React Conf, and support the ecosystem.
- The governing board includes representatives from Amazon, Callstack, Expo, Meta, Microsoft, Software Mansion, and Vercel; Meta commits over $3 million and engineering support for five years.
FleetCode: Multi-Agent CLI Coding in Isolated Git Worktrees
FleetCode is a lightweight desktop app that runs multiple CLI coding agents in isolated git worktrees, supporting persistent sessions, MCP server management, and terminal theming.
- FleetCode is a desktop application enabling parallel execution of CLI coding agents (Claude Code, Codex) within isolated git worktrees
- Supports multiple sessions, persistent across restarts, with features like session renaming, closing, and worktree cleanup
- Requires Node.js 16+, Git, and Claude CLI (
npm install -g @anthropic-ai/claude-cli
); offers commands for development (npm run dev
) and production (npm run build
,npm start
)
DIY ESP32 One-Handed Chorded Keyboard with Low Latency and Long Battery Life
Firmware and design guides for a DIY one-handed chorded keyboard built with ESP32, featuring optimized layouts, low latency, long battery life, and easy construction using modeling clay and off-the-shelf components.
- Firmware and accessories for creating a one-handed chorded keyboard (KEYER) using ESP32-based LilyGO T-Energy S3 board
- Features include minimal finger movement, hand-free operation, attachment to gloves, 215+ chords, arpeggios, multiple layers, optimized ergonomic layout, low latency, months-long battery life, and easy DIY assembly with modeling clay
- Building materials include LilyGO T-Energy S3 ($9.70), Samsung INR18650-35E battery ($2.95), modeling clay ($2.75), Gateron G Pro 3.0 switches (
$10), keycaps ($8), and copper wire (~$1); total cost approximately $34.40
Locket Launches Rollcall Weekly Photo Dumps for Gen Alpha Users
Locket introduces Rollcall, a weekly photo dump feature for iOS and Android, targeting Gen Alpha users, with 80% of weekly posters being this demographic, to enhance private friend-to-friend sharing.
- Locket is launching the Rollcall photo dump feature in October, enabling users to upload weekly collections viewable for seven days.
- The feature has been tested in the US for six months and will be available on iOS next week, with Android rollout later this year.
- About 80% of users posting weekly photo dumps are Gen Alpha (ages 13-17), comprising 49% of US users; Locket has 9 million global daily active users and over 90 million downloads.
Recall v1.6.0 Adds Secure Redis Storage and Advanced Memory Management for Claude
@joseairosa/recall v1.6.0 provides a Redis-backed memory extension for Claude, enabling persistent, organized context storage with security considerations, supporting cross-session recall, relationships, versioning, and advanced management features.
- @joseairosa/recall v1.6.0 published 16 hours ago, enabling Claude to store conversation memories in Redis with persistent memory across sessions and context limits
- Stores sensitive data including code snippets, API keys, business logic, personal info, and organizational details, requiring Redis security best practices
- Implements memory organization by workspace, with options for workspace modes (
isolated
,global
,hybrid
) and features like memory relationships, version history, templates, categories, TTL, export/import, duplicate detection, and analytics
Intel Plans to Restrict Open Source Contributions to Protect Proprietary Advantage
Intel’s leadership signals a shift toward restricting open source contributions to protect proprietary advantages amid resource constraints, risking community fragmentation and reduced support.
- Kevork Kechichian, head of Intel’s datacenter division, advocates rethinking open source contributions to prioritize benefits for Intel.
- Kechichian emphasizes the need to prevent competitors from gaining more from Intel’s open source efforts.
- Intel commits to maintaining open source involvement but considers restricting access to certain codebases, such as low-level libraries, to protect proprietary advantages.
- Intel’s open source contributions include libraries like OneMKL, which have implemented platform detection and restrictions to limit rival optimization.
- The company faces resource constraints due to layoffs, leading to orphaned Linux packages and reduced maintenance of key drivers and libraries.
- Intel aims to enhance its open source strategy to reinforce community support while safeguarding its competitive edge.
Meta Launches React Foundation to Promote Open Governance and Reduce Vendor Control
Meta will contribute React, React Native, and JSX to a Linux Foundation-backed React Foundation with independent governance to address ecosystem fragmentation and vendor dominance concerns.
- Meta will transfer React, React Native, and JSX to a new React Foundation under the Linux Foundation to prevent vendor dominance.
- The foundation will have seven initial members: Amazon, Callstack, Expo, Meta, Microsoft, Software Mansion, and Vercel; Seth Webster, Meta’s React head, will be the first executive director.
- React, created by Facebook in 2013 and used by over 80% of developers according to the 2024 State of JavaScript, is not a full framework and lacks features like data fetching and routing.
▶️ Software Development
Google Launches Gemini CLI Extensions for Customizable Developer Workflows
Google introduces Gemini CLI extensions, enabling developers to customize and connect Gemini CLI to external tools, creating personalized, intelligent command-line workflows.
- Gemini CLI extensions framework announced to enable customization and integration of external tools via command line.
- Over one million developers have adopted Gemini CLI since its launch, accessing an ecosystem of extensions from Google, industry partners, and open-source contributors.
- Extensions include pre-packaged integrations with tools like Dynatrace, Elastic, Figma, Harness, Postman, Shopify, Snyk, and Stripe, each with built-in “playbooks” for immediate use.
- Installation is simplified with the command:
gemini extensions install <GitHub URL or local path>
. - Extensions can bundle MCP servers, context files, custom commands, and exclude tools, enabling personalized, intelligent workflows.
- Google has developed extensions for cloud-native deployment, GKE management, Google Cloud interaction, code review, security, Maps, Flutter, Chrome DevTools, Firebase, and Genkit.
- Extensions enhance Gemini CLI’s intelligence by wrapping MCP tool connections with layers of personalization and context-aware execution.
- A new Gemini CLI Extensions page hosts a catalog of community, partner, and Google-built extensions, ranked by GitHub stars.
Lavandula: A Fast Lightweight C Web Framework for Modern Development
Lavandula is a fast, lightweight C web framework with routing, middleware, SQLite, JSON, and testing, aiming for simplicity and performance; ongoing development adds JSON parsing, ORM, and templating.
- Lavandula is a lightweight, fast web framework written in C, designed for modern web application development.
- Features include routing, HTTP method support, middleware pipeline, SQLite and JSON integration, built-in logging, environment variables, and unit testing.
- In-progress features include JSON body parsing, session cookies, CORS policies, Lavandula ORM, and embedded HTML templating; future plans include rate limiting, static file serving, database integrations, dependency injection, and route listing.
LLM Coding Agents Face Copy-Paste and Clarification Challenges
LLM coding agents struggle with copy-paste operations and asking questions, relying on memory-based write
commands and assumptions, hindering their effectiveness as coding assistants.
- LLM coding agents lack true copy-paste capabilities; they rely on
write
commands from memory, making code movement feel unnatural - They are poor at asking clarifying questions, often making assumptions and brute-forcing solutions instead of engaging in interactive problem-solving
- These limitations contribute to the perception that LLMs are more like overconfident interns than reliable developers
▶️ Management and Leadership
n8n Raises $180M in Series C to Advance AI Orchestration and Business Integration
n8n secured $180 million in Series C funding to enhance AI orchestration and integration, aiming to improve reliable deployment of AI in business processes, with a focus on flexibility and community-driven development.
- n8n raised $180 million in Series C funding, totaling $240 million with a valuation of $2.5 billion
- Funding led by Accel, with support from Meritech, Redpoint, Evantic, Visionaries Club, NVIDIA’s NVentures, T.Capital, Felicis Ventures, Sequoia, Highland Europe, and HV Capital
- Investment emphasizes the importance of reliable AI deployment within business workflows, focusing on orchestration and coordination of AI agents
German Ministers Oppose Suspicionless Chat Monitoring at EU Level
German officials, including Minister of Justice Stefanie Hubig and Patrick Breyer, oppose suspicionless chat control proposals at the EU level, affirming privacy rights and legal safeguards.
- German Federal Minister of Justice, Dr. Stefanie Hubig, stated that “suspicionless #Chatkontrolle must be taboo in a Rechtsstaat” on October 8, 2025
- Germany’s Minister of Justice, Patrick Breyer, emphasized that “private communication cannot be placed under general suspicion” and that Germany will oppose such proposals at the EU level (source)
- The statements oppose EU-wide legislation for suspicionless chat monitoring, emphasizing protection of privacy and legal standards in Germany and the EU
Managed Email Provider Uses Surveillance and Lock-In Tactics to Trap Clients
A managed email service provider used contractual clauses and potential email surveillance to prevent client exit, exemplifying vendor lock-in and unethical practices in open source software deployment.
- The article narrates a real-world horror story highlighting vendor lock-in, contractual manipulation, and surveillance tactics used by a managed service provider for open source email solutions.
- The provider extended notice periods from 6 to 12 months, introduced paid features, and disabled alternative access methods, increasing costs by 30% and restricting data ownership.
- Evidence suggests the provider may have read customer emails, as they threatened legal action and cited internal details, possibly accessing messages via email client tokens.
Secure Your Email: Own Your Domain and Build Resilient Backup Systems
The article advocates for owning a custom domain and setting up backup email infrastructure, including local or self-hosted servers, to maintain unbannable, resilient email access and control over critical digital communications.
- The article emphasizes the importance of owning and controlling your email domain to prevent bans and ensure access.
- Recommends using providers like Google Workspace, Proton Email, or Migadu for domain-based email setup with DNS MX records and catch-all features.
- Suggests establishing a local or self-hosted email server (e.g., on NAS or via OpenBSD) to create a backup and maintain access if provider accounts are banned, with forwarding setup to ensure continuous email receipt and resilience against bans.
Cory Doctorow Warns Big Tech Uses Apps to Evade Regulation and Foster Monopolies
Cory Doctorow’s Enshittification (2025) argues that Big Tech exploits app-based legal loopholes and industry concentration to bypass regulation, fostering monopolies and regulatory capture.
- Cory Doctorow’s book Enshittification (2025) critiques how Big Tech uses apps to evade regulation and facilitate monopolistic practices.
- Tech companies manipulate legal frameworks by claiming crimes are committed “with an app,” exploiting differences in legal regimes for apps versus web pages.
- Concentration of industries into few firms (cartels) reduces regulatory oversight, enabling regulatory capture and anti-competitive behaviors.
Financial Times Subscription Page Promotes Digital and Print Plans
The article is a subscription landing page for the Financial Times, offering various digital and print plans without providing specific news or technical information.
- Article titled “Subscribe to read” is a placeholder page with no substantive content.
- Contains navigation, subscription offers, and promotional links for FT digital and print products.
- No technical data, metrics, or key developments are presented in the article content.
California Enacts First U.S. Law for Easy Browser Data Opt-Out
California’s new law mandates browsers to enable a simple, universal opt-out mechanism for third-party data sharing, significantly improving user privacy control and building on the 2018 California Consumer Privacy Act.
- California Gov. Gavin Newsom signed a bill requiring browsers to provide an easy-to-find, one-click mechanism for Californians to opt-out of third-party data sales.
- The law enhances the 2018 California Consumer Privacy Act by simplifying universal opt-out signals, replacing the need for third-party extensions or privacy browsers.
- The bill is the first of its kind in the U.S., passed on September 11, and mandates that browsers implement accessible opt-out features for all users.
- Additional bills signed include mandates for social media account deletion with full data removal and increased transparency from data brokers regarding personal data collection.
OpenAI Warns EU on Big Tech Data Dominance and Market Lock-In
OpenAI warned EU antitrust authorities about Big Tech’s data dominance, citing market barriers and urging intervention to prevent customer lock-in by Google, Microsoft, and Apple.
- OpenAI raised concerns with EU antitrust enforcers about dominant Big Tech companies like Google, Microsoft, and Apple.
- The company highlighted “difficulties” in competing with entrenched firms and urged intervention to prevent customer lock-in by large platforms.
- The meeting with EU antitrust chief Teresa Ribera occurred on September 24, emphasizing risks of market entrenchment and anti-competitive practices.
OpenAI and Nvidia Announce $100B Data Center Investment and AMD Partnership
OpenAI’s $100 billion data center buildout with Nvidia chips and a partnership with AMD for chip deployment create interconnected deals fueling the $1 trillion AI market.
- OpenAI and Nvidia announced a $100 billion investment deal to build data centers with Nvidia’s advanced chips.
- OpenAI committed to filling these centers with millions of Nvidia chips, fueling a circular financing structure.
- Separately, OpenAI signed a partnership with AMD to deploy tens of billions of dollars’ worth of AMD chips, with OpenAI becoming one of AMD’s largest shareholders.
Google Revises Health Policy to Ensure Employee Data Privacy and Voluntary AI Tool Use
Google revised its internal health policy following employee concerns, making participation in Nayya’s AI health tool optional and ensuring no data sharing occurs without consent.
- Google updated internal health policy after employee backlash regarding third-party AI health tool access
- Initially required employees to share personal data with Nayya to qualify for benefits; policy clarified to make opt-in voluntary
- No data will be shared with Nayya for employees who do not opt-in; pay, gender, and social security number shared only if opted in
IMF and BoE Warn Rapid AI Growth Could Trigger Market Crash
IMF and BoE warn that rapid AI growth could trigger an abrupt stock market correction, posing risks to financial stability and emphasizing the need for regulatory measures.
- IMF and BoE warn that AI boom risks causing an “abrupt” stock market correction.
- The warning emphasizes potential market volatility due to rapid AI development.
- The article highlights concerns over AI’s impact on financial stability and the need for regulatory oversight.
LayerX Report Reveals 77% of Employees Exfiltrate PII into AI Chatbots
LayerX’s 2025 report reveals widespread enterprise AI data exfiltration, with 77% of employees copying data—including PII/PCI—into ChatGPT, often from unmanaged accounts, raising security and compliance concerns.
- Study by LayerX reports 45% of enterprise employees use generative AI tools, with 77% copying and pasting data into chatbots; 22% of these include PII/PCI.
- 82% of pasted data originates from unmanaged personal accounts, creating significant blind spots for data leakage and compliance risks.
- About 40% of file uploads to AI sites contain PII/PCI data, with 39% from non-corporate accounts; LayerX monitors data via browser extensions, not API calls.
OpenAI Bans Chinese and Russian Accounts for Surveillance and Malware Planning
OpenAI banned suspected Chinese and Russian accounts for misuse, including surveillance planning, influence campaigns, and malware development, using ChatGPT and multiple AI models; over 40 networks have been banned since February 2024.
- OpenAI banned ChatGPT accounts linked to Chinese government entities for attempting to use AI models to plan surveillance activities.
- The banned users typically asked ChatGPT to help design tools for large-scale monitoring, analysis, influence campaigns, and malware development.
- Since February 2024, OpenAI has banned over 40 networks violating usage policies, with threat actors using multiple AI models for automation and evasion.
Virgin Media O2 Launches Free Scam Schools to Protect Vulnerable UK Adults
Virgin Media O2’s Scam Schools initiative provides free UK-wide fraud prevention education for vulnerable and older people, addressing rising scam victimization amid bank closures and digital dependence.
- Virgin Media O2 launched the “Scam Schools” program in October 2025, partnering with Good Things Foundation to educate vulnerable groups on online fraud prevention.
- The initiative offers free, in-person sessions across the UK, with registration available via the website and online application forms.
- The program aims to combat rising online scams targeting older adults, with over 1.8 million over-65s in the UK scammed in the past year, averaging £831 loss per victim, amid increased bank branch closures and online reliance.
Millions of Business PCs Face Windows 10 Support End Amid Hardware Challenges
Hundreds of millions of business PCs remain on Windows 10 as support ends October 14, 2025, with many lacking hardware for Windows 11; cost of ESUs increases, prompting hardware upgrades.
- Microsoft will end Windows 10 support on October 14, 2025, offering extended security updates (ESUs) at $61 for 12 months, doubling annually.
- Approximately 550 million Windows devices are in corporate environments; about half may not meet Windows 11 hardware requirements.
- Omdia estimates 20% of non-upgrading machines lack Trusted Platform Module (TPM) or modern CPU hardware; global Windows 11 adoption is 47-50%, higher in US (60%) and EMEA (65%).
Gartner Warns AI Market Oversupply Will Lead to Consolidation and Correction
Gartner predicts market correction due to oversupply of agentic AI, favoring financially robust incumbents; the correction is driven by economic realities, not systemic bubbles, with long-term consolidation expected.
- Gartner warns that oversupply of agentic AI models, platforms, and products will lead to market consolidation and correction.
- The market shift is driven by fading hype and FOMO, with economic fundamentals becoming more prominent.
- Leading vendors like OpenAI, Nvidia, AMD, and Oracle are investing approximately $500 billion annually until 2030 to develop AI infrastructure.
McKinsey Warns AI Vendors May Hike Prices Without Delivering ROI
McKinsey warns AI software vendors may hike prices without delivering measurable benefits, citing ROI issues, high deployment costs, scaling challenges, and evolving pricing models.
- McKinsey’s report highlights that 70% of AI use cases lack published quantifiable ROI.
- Software vendors risk increasing prices without reducing costs or improving productivity, potentially inflating AI deployment costs by 60-80%.
- Challenges include scaling adoption due to underinvestment in change management, unpredictable pricing models, and the need for hybrid consumption-based pricing.
▶️ Technology
Cloudflare Finds and Fixes Race Condition in Go’s arm64 Compiler
Cloudflare identified a race condition in Go’s arm64 compiler related to split stack adjustments, fixed in Go 1.23.12+ by making stack pointer updates atomic to prevent preemption during critical instructions.
- Cloudflare discovered a bug in Go’s arm64 compiler causing a race condition during stack unwinding, linked to split stack pointer adjustments.
- The bug manifests when preemption occurs between two specific assembly instructions in the function epilogue, leading to invalid stack pointers and crashes.
- The issue was fixed in Go versions 1.23.12, 1.24.6, and 1.25.0 by modifying the compiler to generate atomic stack pointer updates, eliminating the race window.
Python 3.14 Boosts Performance by 21% as Fastest CPython Version
Python 3.14, released on October 8, 2025, improves performance by about 21% over 3.13, with benchmarks indicating it is the fastest CPython version for pure Python code; JIT does not significantly enhance speed, while free-threading offers notable multi-threaded gains.
- Python 3.14 was officially released on October 8, 2025, with benchmark tests conducted immediately afterward.
- Benchmarks focused on pure Python scripts (recursive Fibonacci and bubble sort) across multiple Python versions, interpreters, threading modes, and hardware platforms.
- Results show Python 3.14 is approximately 21% faster than Python 3.13 in single-threaded Fibonacci, with the fastest overall performance among CPython versions tested.
Reilly Chase Builds Poker Stats Web App in Days Using Cursor AI
Reilly Chase built a comprehensive poker stats web app with Cursor AI, automating features like hand parsing, balance checks, and analytics, entirely through conversational prompts over 2-3 days.
- Built a web app at poker.rchase.com using Cursor AI in 2-3 days through conversational prompts
- Features include admin dashboard, PokerStars and Gmail integrations, over 700 lines of hand history parser, poker stats calculations (VPIP, PFR, 3-Bet), CRUD for journal entries, multi-file upload, daily balance checks, auto export/import, hand viewer, profit/loss chart
- Deployed on DigitalOcean Debian 13 server, developed locally with Herd for MacOS, all code generated via Cursor AI conversations without manual coding
Svelte Tops JavaScript Frameworks in Performance, Study Finds
A 2021 academic study finds Svelte is the fastest JavaScript framework, outperforming Angular, React, Vue, and Blazor in creation and update benchmarks due to its reactive system and compiler optimizations.
- A 2021 study compares Angular, React, Vue, Svelte, and Blazor, published in the Journal of Web Engineering
- Svelte outperforms other frameworks in creation and update benchmarks, leveraging reactive system, compiler optimizations, and binding-based rendering
- Blazor, using WebAssembly, shows significantly worse performance than JavaScript-based frameworks
Google Cloud Launches Gemini Enterprise AI for Workplace Automation
Google Cloud introduced Gemini Enterprise, a workplace AI platform for automating tasks and content creation, priced at $30 per user monthly, competing with Microsoft and OpenAI.
- Google Cloud launched Gemini Enterprise, an AI platform targeting workplace automation, with a $30/month per user fee.
- Aims to enable employees to automate tasks and generate content across departments.
- Unveiled ahead of the Gemini at Work event, competing with Microsoft and OpenAI in enterprise AI tools.
ChatGPT Surpasses 800 Million Weekly Users and 700 Million Monthly Active Users
ChatGPT has rapidly expanded to 800 million weekly users and 700 million monthly users, processing over 2.5 billion messages daily, demonstrating unprecedented global diffusion and dominance in AI.
- ChatGPT is used by 10% of the world’s adult population, approximately 700 million monthly users as of July 2025
- Weekly active users reached 800 million, with over 2.5 billion messages processed daily (around 29,000 queries/sec)
- User growth has increased 700% since November 2023, when it had 100 million active users, surpassing other LLM chatbots like xAI’s Grok (65 million) and Anthropic’s Claude (30 million)
Samsung S90D OLED 4K TV Combines QD-OLED Brightness and Gaming Features
Samsung’s S90D OLED 4K TV (2024) combines QD-OLED technology with Tizen OS, offering 1,240 nits brightness, four HDMI 2.1 ports, and advanced gaming features, with some interface and HDR format limitations.
- The Samsung S90D OLED TV features a QD-OLED panel with 4K resolution, 144Hz refresh rate, and four HDMI 2.1 ports, delivering peak brightness of 1,240 nits.
- The TV uses Tizen OS, supports HDR10, HDR10+, HLG, and Dolby Vision, and includes a gaming hub with cloud services like Xbox Game Pass.
- The S90D offers pixel-level contrast, wide viewing angles, and vivid colors, but has interface quirks, limited Dolby Vision support, and no 83-inch size option.
Jony Ive Plans 15-20 Human-Centric AI Devices for OpenAI
Jony Ive revealed he has 15-20 innovative AI device ideas for OpenAI, aiming to create fun, human-centric products that address issues caused by smartphones and tablets.
- Jony Ive has 15 to 20 ideas for OpenAI’s family of AI devices, likely distinct from iPhone-like products
- Ive criticized smartphones and tablets, emphasizing designs that improve human happiness and reduce anxiety
- OpenAI acquired Ive’s startup for over $6 billion earlier this year, fueling speculation on new AI gadgets
OpenAI Unveils ChatGPT App Store with Major Partners and Future Developer Access
OpenAI launched an app store within ChatGPT, enabling users to interact with apps via natural language, with developer submissions opening in late 2025 and monetization through integrated commerce.
- OpenAI announced a new app store integrated within ChatGPT, with apps like Spotify and Zillow available.
- Developers can submit apps later in 2025, with monetization options including OpenAI’s new commerce functionality.
- The app store features a directory with curated apps, and initial partners include Booking.com, Canva, Coursera, Figma, Expedia, Spotify, and Zillow.
OpenAI’s Sora Hits 1 Million Downloads Amid Copyright Concerns
OpenAI’s Sora app achieved 1 million downloads within five days of launch, enabling free AI-generated short videos on iOS, amid copyright infringement criticism from industry groups.
- OpenAI’s Sora, a short-form AI video app, reached 1 million downloads in less than five days post-launch in late September.
- Sora allows users to generate short videos for free by typing prompts and is available exclusively on iOS via invite-only access.
- Despite rapid growth and reaching No. 1 in Apple’s App Store, Sora faces backlash from the Motion Picture Association over copyright infringement concerns.
Figure 03: The Next-Gen Humanoid Robot with Advanced Sensing and Mass Production
Figure 03, introduced on October 9, 2025, is a scalable, mass-manufacturable humanoid robot with advanced perception, tactile sensing, safety features, and wireless charging, enabling broad domestic and commercial applications.
- Figure 03 is the third-generation humanoid robot, introduced on October 9, 2025, designed for Helix, home, and global deployment.
- Features a redesigned sensory suite with high-frequency vision cameras delivering twice the frame rate, quarter latency, and 60% wider field of view; embedded palm cameras for occlusion resilience; tactile fingertips with force detection as low as 3 grams; and 10 Gbps mmWave data offload.
- Incorporates safety enhancements including multi-density foam, soft textiles, washable removable soft goods, wireless inductive charging at 2 kW, and a new high-volume manufacturing process via BotQ, capable of producing 12,000 units annually with a goal of 100,000 over four years.
Microsoft Closes Windows 11 Loophole for Account-Free Installations
Microsoft closed a loophole in Windows 11 Insider builds that allowed installation without a Microsoft account, requiring users to complete OOBE with internet and account verification.
- Microsoft has introduced a fix in recent Windows 11 Insider builds (Build 26220.6772) to close a loophole allowing installation without a Microsoft account.
- The workaround involved executing
start ms-cxh:localonly
, which is now disabled, as Microsoft states these mechanisms can cause incomplete device setup. - Users previously bypassed account creation during OOBE, risking incomplete configuration; Microsoft emphasizes that completing OOBE with internet and a Microsoft account is now mandatory for proper setup.
Widespread AI Use Among Students Raises Concerns Over Social Skills and Safety
A CDT study warns widespread AI use among students may impair social skills and critical thinking, with risks including data breaches, harmful interactions, and mental health issues, amid insufficient teacher training.
- A survey by the Center for Democracy and Technology found 85% of teachers and 86% of students used AI in the past year, with only half of students using AI for schoolwork.
- AI use correlates with increased data breaches, sexual harassment, bullying, AI failures, and troubling student interactions; 42% of students used AI for mental health, companionship, or escape, and 19% formed romantic relationships with chatbots.
- Teachers express concern that AI diminishes social skills, critical thinking, and writing abilities, with MIT research showing reduced brain activity and poorer recall in students using AI for writing.
Humans Struggle to Differentiate AI-Cloned Voices from Real Speech
Researchers found humans cannot reliably distinguish AI-generated voices from real ones, with cloned voices reaching near-human realism using minimal sample data, complicating detection efforts.
- Study by Queen Mary University of London and University College London found humans cannot reliably distinguish real from AI-cloned voices, with detection accuracy near chance.
- Tested 80 audio samples (half human, half synthetic); synthetic voices generated with less than five minutes of sample speech using off-the-shelf software from ElevenLabs.
- Fully synthetic AI voices were identified as human only 41% of the time; cloned voices from real speech fooled listeners 58% of the time, with no significant sensitivity.
- Participants correctly identified real voices 62% of the time; results indicate that cloned voices can reach a human-like level of realism, making detection difficult.
- Findings suggest that advanced AI voice cloning, using limited input data, can produce voices indistinguishable from real recordings, raising concerns over scams and misinformation.
Microsoft Windows 11 Insider Preview Adds Redesigned Start Menu and New Features
Microsoft’s Windows 11 Canary build 27965 introduces a redesigned Start menu, removes .NET 3.5 SP1, and adds the command-line editor Edit, enhancing user interface and legacy support.
- Microsoft has released Windows 11 Insider Preview build 27965 to Canary Channel testers, featuring a redesigned Start menu.
- The new Start menu includes a category view, application grouping, and Phone Link integration; EEA users will access iOS and Android options later in 2025.
- Build 27965 removes .NET Framework 3.5 SP1 from Features On Demand and introduces the command-line text editor, Edit.
Intel’s Panther Lake Core Ultra Series 3 to Launch in 2025 with Major GPU Gains
Intel’s Panther Lake (Core Ultra Series 3), arriving in late 2025/early 2026, offers up to 16 cores, 12 Xe3 GPU cores, 50% GPU performance increase, and improved power efficiency on the 18A process.
- Intel’s Panther Lake, aka “Core Ultra Series 3,” scheduled for late 2025/early 2026 release, built on Intel’s 18A process, with production on Intel’s Arizona Fab 52.
- Features 8- and 16-core CPUs with four Xe3 graphics cores, and a 16-core CPU with 12 Xe3 cores and 12 ray-tracing units—highest integrated GPU horsepower to date.
- Promises up to 10% lower power consumption than Lunar Lake, with 40% lower power at similar single-threaded performance and 50% more GPU power over predecessors; supports up to 96GB LPDDR5 or 128GB DDR5 memory, includes AI NPU, updated image processing, and 10-bit AVC/AV1 video encoding.
WinBoat Brings Seamless Windows App Support to Linux with USB and Filesystem Features
WinBoat (release v0.8.7) provides seamless Windows app integration on Linux, supporting automated installs, filesystem sharing, USB passthrough, and running Office 365, with plans for GPU and Podman support.
- WinBoat enables running Windows applications seamlessly on Linux with an integrated interface, automated installation, and filesystem sharing.
- Supports most Linux distributions via AppImage; current version is v0.8.7 in beta, with features like USB passthrough (from v0.8.0), but no GPU passthrough yet.
- Allows running Windows apps including Office 365, with plans for Podman support and potential Flatpak integration; does not support kernel anti-cheat games or GPU acceleration currently.