Mamdani Accuses New York of Hiding 9/11 Air Toxicity Data from Public

New York City mayoral candidate Zohran Mamdani has alleged that city authorities knowingly withheld information about toxic air quality following the September 11, 2001 attacks. The attacks, which killed 2,977 people primarily due to the collapse of the Twin Towers, left behind hazardous air conditions in lower Manhattan. Mamdani's claims suggest officials were aware of the health risks but chose not to disclose them to the public. The allegation has surfaced ahead of the 25th anniversary of the 9/11 attacks.

3D-Printed Eggshell Arrays Filled With Water Boost Spacecraft Impact Resistance

Researchers have developed a novel shielding material inspired by eggshell structures to better protect next-generation spacecraft. The design involves 3D-printed aluminum arrays modeled after eggshells, which are then filled with water. Testing showed this approach outperformed conventional shielding methods in resisting high-velocity impacts. The innovation could play a significant role in safeguarding future spacecraft from debris and micrometeorite collisions.

Developer Proposes Two-Layer Architecture to Limit AI Agent API Risks

A developer experimenting with agent safety has identified a risky pattern in modern AI agent setups, where language models are handed API tokens and can directly execute backend calls without independent oversight. The researcher proposes separating reasoning and execution into two distinct layers: a probabilistic 'Semantic Plane' where AI models interpret intent but hold no credentials, and a deterministic 'Control Plane' built in Rust that validates proposed actions against hard policies before granting execution rights. The Control Plane issues short-lived, single-use cryptographic grants, ensuring no action reaches a backend service without passing structured checks. The approach aims to prevent issues such as prompt injection, hallucination-driven misfires, and runaway recursive calls between collaborating agents. The author stresses this remains an early-stage personal exploration and has not been formally audited or proposed as an industry standard.

Google Cloud Partners with Accenture to Accelerate Enterprise AI Deployment

Google Cloud has announced a strategic partnership with Accenture aimed at expanding enterprise AI adoption. The deal focuses on deploying forward-deployed engineers to help businesses implement AI solutions more effectively. The collaboration is designed to address common bottlenecks that slow down AI deployment in large organizations. Google Cloud is using this partnership to strengthen its competitive position in the rapidly growing enterprise AI market.

Five Juvenile Orangutans Found Abandoned in Odisha Forest, Trafficking Suspected

Five young orangutans were discovered abandoned in a forest in Odisha's Balasore district after local villagers spotted them in the trees and notified authorities. Forest officials believe the animals were brought into India through an international wildlife trafficking network. Forest personnel managed to safely recover all five apes without the use of tranquilizers. The Wildlife Crime Control Bureau has launched an investigation to determine how the orangutans entered the country and trace their origin.

Do you think it happened? Research stolen from their Codex private chats

Article URL: https://www.reddit.com/r/codex/s/psWHRr9864 Comments URL: https://news.ycombinator.com/item?id=49612463 Points: 12 # Comments: 2

MyZubster Metaverse Upgrades Realtime Infrastructure With Observability and Durability

The MyZubster Metaverse platform has merged a significant update to its realtime infrastructure, introducing a dedicated observability layer for its Socket.IO gateway that tracks connection metrics, message delivery, latency, and service-level objectives. The system is designed to be privacy-safe, using hashed references instead of raw user identifiers in logs and metrics. A bounded operation queue has been added to manage traffic bursts, returning retryable responses to clients when the system is under load rather than allowing unbounded memory accumulation. Message and notification persistence has been improved so that a failed socket delivery no longer discards data that was already successfully stored. The update also includes Redis failure handling with a local fallback, an authenticated admin metrics endpoint, and a test suite covering load bursts, durability, and authorization scenarios.

GEO and AEO Audits Now Essential as AI Crawlers Operate Beyond Google Search Console

Websites that appear fully healthy in traditional SEO reporting may still be inaccessible or unreadable to AI-powered search systems, creating a blind spot for publishers. Google Search Console remains a valuable tool for understanding how Google indexes pages, but it does not reveal whether AI crawlers can access a site or interpret its content. Industry reporting, including guidance from Search Engine Land, highlights that hosting configurations and crawler-specific rules can silently block AI crawlers even when conventional indexing looks clean. An emerging five-layer GEO measurement framework calls for audits that go beyond standard dashboards to check AI-crawler access, entity clarity, and citation signals in AI-generated answers. Experts now consider Generative Engine Optimization and Answer Engine Optimization core components of modern search visibility strategy rather than optional add-ons.

Airport chaos in UK: Nearly 300 flights cancelled due to air traffic control glitch

Laravel 13 Adds Native Vector Search for MariaDB, MySQL Users Still Excluded

A pull request merged into Laravel's 13.x branch on August 20 extended the framework's native vector search capabilities — including methods like whereVectorSimilarTo and orderByVectorDistance — to MariaDB, after previously supporting only PostgreSQL with pgvector. The update refactored vector distance logic from a hardcoded PostgreSQL check in the query builder into driver-level grammar methods, making it easier to add support for additional databases. MariaDB gains this support because it ships a native vec_distance_cosine() function and VECTOR column type since version 11.7 Community. Standard MySQL installations remain unsupported because MySQL lacks a native vector distance function outside of Oracle's HeatWave and MySQL AI cloud offerings. Notably, the Laravel team chose to throw a clear RuntimeException rather than implement a silent in-memory PHP fallback, which would have broken pagination semantics and query performance guarantees.

Google's 30-Day Gemini Preview Shutdown Exposes SaaS Model Dependency Risks

Google announced on August 27 that its Gemini Omni 1.1 Flash video generation model would move to general availability, simultaneously scheduling the older preview endpoint for deprecation on September 30 — giving developers roughly 30 days to migrate. The author, building a salon AI app called Miraviso, notes this pattern is a systemic risk for any product built on hosted AI model previews. Miraviso's on-device color try-on feature is unaffected since it runs locally via MediaPipe, but its server-side haircut preview on Vertex AI faces a real dependency on Google's release schedule. A privacy-first design — where preview images are never stored — removes the standard migration tool of replaying production data against the new model. The author argues that the difference between a stressful migration and a smooth one is whether a tested procedure already exists before the deprecation notice arrives.

How Sharing Your Startup's Build Process Can Replace Paid Marketing

Building in public involves sharing real development decisions, fixes, and progress openly, turning everyday work into discoverable content without ad spend. Founders can document changes such as onboarding redesigns or bug fixes as short posts, giving potential users and peers concrete insight into how a product evolves. Early and ongoing public updates help audiences build context about a product before a formal launch, making announcements more meaningful to followers. Public posts can also generate feedback that directly improves the product, as audience questions may reveal missing features or unclear messaging. To stay sustainable, founders are advised to keep the habit small and focused, sharing one meaningful update at a time while tracking whether conversations attract genuinely relevant audiences.

Wrong USB Cable Silently Caps Tether Speed to 480 Mbps, Even on 10 Gb/s Phones

A developer investigating slow USB tethering discovered that using a USB 2.0 cable forces the entire link to negotiate at 480 Mbps, regardless of what the phone or laptop supports. USB link speed is jointly negotiated between the host controller, the cable, and the device, with the slowest component setting the ceiling. Crucially, no system warning or error is raised when a cable downgrades the connection — the only evidence is a value buried in sysfs. Testing across 45 measurement records and 14 handsets found that 41 out of 45 sessions enumerated at 480 Mbps, including an iPhone 17 Pro rated at 10 Gb/s. Switching to a SuperSpeed cable dropped bus utilisation from 91% to just 3.6% on the same link, illustrating the dramatic real-world headroom difference.

Open-Source Terraform Project Deploys Full AWS Production Stack in One Command

A GitHub project called terraform-aws-infrastructure, created by manishpcp, offers a fully modular Terraform setup that provisions an entire production-grade AWS environment with a single command. The project addresses common infrastructure pain points such as configuration drift, overexposed compute, manual TLS certificate management, and absent observability. The stack spans two Availability Zones and includes a private-subnet EC2 setup behind an Application Load Balancer, CloudFront-backed static hosting, a containerized Flask microservice on ECR, and automated DNS via Route 53. CI/CD pipelines authenticate through OIDC rather than stored AWS credentials, eliminating long-lived secrets from GitHub Actions. Built with Terraform 1.5.0 or later, the project uses a clean modular layout with remote state managed via S3 and DynamoDB.

How to Safely Remove Login Methods on Multi-Identity Account Pages

A technical guide published on DEV Community outlines best practices for managing and removing login methods on multi-identity account pages, particularly in fintech applications. The core principle is to treat login methods as recoverable credentials, meaning a method should only be removed after verifying the user retains an alternate recovery path. The process requires step-up authentication before removal and immediate revocation of all sessions tied to the deleted identity. A server-side API should own the authoritative state of each login method, with the UI rendering controls based on those facts and re-validating them at the point of action. The guide also warns against race conditions — such as two browser tabs simultaneously removing the last two methods — and recommends database transaction locks to prevent accidental account lockouts.

CSS Anchor Positioning Enables Smooth Grid Reflow Animations Without JavaScript

A technique using CSS anchor positioning allows auto-fill grid cells to animate smoothly during reflow, something previously impossible because grid cells lack interpolatable positions. Each cell gets a wrapper element assigned as a named anchor, while the visible content inside is absolutely positioned and stretches to match it using inset: anchor(inside). When the grid recomputes track layout, the browser interpolates the inset values over a plain CSS transition, making cells appear to glide rather than snap to their new positions. The approach requires no JavaScript, no ResizeObserver, and no View Transitions, and unlike View Transitions it remains interruptible mid-animation. Firefox currently does not support the transition, so a @supports feature-detect is recommended to gracefully fall back to the standard instant reflow behavior.

GrapheneOS and Motorola Partnership Draws Skepticism From Privacy Community

GrapheneOS, a privacy-focused mobile operating system, has announced a partnership with Motorola, raising concerns among some users and commentators. A podcast episode from the Switched to Linux channel outlines reasons why the collaboration may not be as beneficial as it appears. The discussion reflects broader skepticism about whether a mainstream hardware manufacturer can align with GrapheneOS's strict privacy and security principles. The post gained attention on Hacker News, though it attracted minimal community discussion at the time of publication.

Tesla Cybercab Has Hidden Virtual Joystick for Manual Control, Riders Discover

The Tesla Cybercab, which launched in Austin, Texas, was designed without a steering wheel or pedals, making it appear fully autonomous with no manual override. However, within days of its public launch, riders discovered a hidden virtual joystick on the vehicle's central display. The feature, found in the lower-left corner of the screen, includes a toggle for manual driving and appears to be intended exclusively for Tesla employees. At least two videos shared on X have shown Cybercab passengers encountering the concealed control interface. The discovery reveals that the vehicle does have a form of manual control, contrary to its fully driverless design premise.

Free Stars: Children of Infinity Releases Free Combat Demo

A free playable combat demo for Free Stars: Children of Infinity is now available to the public. The demo features 1v1 arcade-style space battles reminiscent of classic early-1990s space games. Players can experience the game's combat mechanics before a full release. The demo offers a nostalgic feel for fans of retro space combat games.

OpenHands 1.0 Launches as Self-Hosted Coding Agent With Built-In Safety Controls

OpenHands, an open-source autonomous coding agent, released version 1.0 featuring a rebuilt architecture based on a Software Agent SDK, production-ready Docker sandboxing, and built-in security policies. The new release splits its codebase into four distinct Python packages, separating core logic, tools, workspace management, and API serving. A key safety feature called ConfirmationPolicy pauses agent execution and awaits human approval before carrying out high-risk tool calls. On the SWE-bench Verified benchmark of 500 real GitHub issues, OpenHands scores approximately 68% with a frontier model backend, outperforming Devin 2.0's reported 45.8%. The release positions OpenHands as a credible self-hosted alternative to commercial coding agents for developers concerned about data privacy and API key exposure.