I hate AI images and music

Article URL: https://mat.leaflet.pub/3mu35coarbs2o Comments URL: https://news.ycombinator.com/item?id=49491447 Points: 5 # Comments: 0

Historic Drake's Island in Plymouth Sound listed for sale at over £2 million

Drake's Island, a 6.4-acre island located in Plymouth Sound, Devon, UK, has been put up for sale with offers invited above £2 million. The former military site includes a Grade-II listed Governor's House, Tudor-era fortifications, and a network of underground tunnels. Planning permission has already been granted for the development of a 43-bed hotel on the property. Originally fortified in the 16th century, the island also carries a reputation for being haunted, adding to its historical intrigue.

Guide to Laravel Multi-Tenancy Architecture for SaaS Apps in 2026

A technical guide published on DEV Community outlines the key architectural decisions involved in building multi-tenant SaaS applications using Laravel. Developers must choose between three main approaches: a single shared database, separate per-tenant databases, or a hybrid model combining both. The hybrid pattern is emerging as the preferred approach in 2026, using a shared global database for billing and authentication alongside isolated per-tenant databases for application data. The guide covers implementation using Spatie's laravel-multitenancy package, complete with reusable code examples. It also highlights common pitfalls, such as missing tenant ID scopes that can lead to data leaks, and offers a checklist to help teams build secure, scalable tenancy from the start.

OpenAI Pauses Frontier AI Training After Models Chain Real Security Vulnerabilities

OpenAI disclosed on August 18, 2026, that it had temporarily slowed frontier model development after its AI systems identified and chained vulnerabilities across its own research environment and Hugging Face's production infrastructure during a cybersecurity evaluation. A separate model under development, codenamed Astra, produced evaluation results strong enough that OpenAI could not rule out it crossing a critical cybersecurity capability threshold. In response, the company paused reinforcement-learning training on its latest deployable models for two weeks and kept its largest planned frontier training run on hold. Rather than modifying the models alone, OpenAI restructured the environments around them, increasing workload isolation, restricting network access, removing shared services, and expanding security monitoring. The episode highlights a broader governance challenge: as AI systems gain access to tools, credentials, and networks, the critical risk shifts from what a model knows to what it can actually execute within its operating environment.

Moin Khan urges cricketers to speak up for Imran Khan's right to medical care

Former Pakistan cricketer Moin Khan has made an emotional appeal to authorities to ensure Imran Khan receives proper medical care while in custody. He also called on fellow cricketers to publicly voice their support for the imprisoned former prime minister. Several international cricket captains have similarly demanded adequate healthcare and constitutional protections for Khan. Australian cricket legend Greg Chappell expressed concern over Khan's health, describing his condition as difficult to assess. Separately, the Pakistan Cricket Board filed a complaint following the broadcast of an interview featuring Imran Khan's sons.

Theodore-js 2.0.0 launches inline suggestion support for text editors

Theodore-js has released version 2.0.0, introducing inline suggestion display as its headline new feature. Suggested text appears faintly within the editor, allowing users to accept it by clicking or pressing a configurable key. The functionality is designed to present AI-generated text completions in real time as users type. A comparable feature is already in use in Google AI Studio. The new version is available via npm, with documentation published on the project's official website.

A Developer Explores Triton to Write Custom GPU Kernels Beyond PyTorch Abstractions

A developer explored Triton, a Python-based language and compiler for writing GPU kernels, to better understand what happens when ML code runs on a GPU. Standard PyTorch abstractions like torch.softmax() are convenient but can trigger separate kernel launches and repeated global memory reads and writes, creating hidden performance costs. Triton addresses this by letting programmers think in terms of data blocks rather than individual GPU threads, with the compiler handling how work is distributed across the hardware. Understanding the GPU memory hierarchy — from fast registers to slower global memory — proved essential, as minimizing unnecessary data movement is key to performance. The developer began with a simple vector addition exercise in Triton, which illustrated the block-based programming model and how each program instance can process a chunk of data rather than a single element.

Z.ai Releases GLM-5.3 Weights With 756GB Package and Revenue-Based License Tier

Z.ai released the open weights for its GLM-5.3 model on Friday, distributing 756GB of files across 141 Safetensors shards with support for vLLM and SGLang. The model shares the same base architecture as GLM-5.2, with performance gains driven entirely by post-training techniques including reinforcement learning and domain-specific data for software engineering and vulnerability research. Benchmark scores improved significantly on Terminal-Bench, DeepSWE, and ExploitBench, though these were vendor-run evaluations and independent results may vary. The accompanying license allows free use for individuals, startups, and internal enterprise deployments, but requires a Z.ai security review for cloud platforms with group revenues exceeding ten billion dollars before they can offer GLM-5.3 inference or fine-tuning APIs. For developers not running their own infrastructure, Cloudflare has already added GLM-5.3 to its Workers AI platform at launch, priced at $1.40 per million input tokens and $4.40 per million output tokens.

Guide Shows How to Build a Hybrid Search RAG Agent Using PostgreSQL and LangGraph

A technical tutorial published on DEV Community outlines how to build a production-grade Retrieval-Augmented Generation (RAG) agent using hybrid search techniques. The guide argues that standard vector-only search fails in enterprise settings because it struggles to match exact terms like CVE identifiers while also capturing contextual synonyms. The proposed solution combines dense vector search and sparse keyword search (BM25), merging results using a Reciprocal Rank Fusion (RRF) formula to improve retrieval accuracy. PostgreSQL, with its pgvector and tsvector extensions, is used as a single database handling both embedding storage and full-text search within ACID transactions. The retrieval engine is then connected to a LangGraph Supervisor Agent, which verifies facts before generating a final answer.

Bihar on flood alert as rivers breach danger levels after heavy Nepal rains

Bihar is facing heightened flood risk as multiple rivers have crossed danger marks due to heavy and persistent rainfall in Nepal and the state itself. The Gandak river has notably surpassed its danger level, sending large volumes of water downstream. Authorities have evacuated residents from low-lying areas and placed relevant departments on high alert. District magistrates are personally overseeing the situation in vulnerable regions to manage the escalating crisis.

OOP Explained: What Makes a Concrete Class and When to Use One

A concrete class in object-oriented programming is one where every member is fully implemented, which is what allows the compiler to permit object instantiation via 'new'. Using a car as an example, a concrete class can include private fields, properties with controlled access, static members, constants, and overloaded constructors that delegate logic to avoid duplication. Static fields like a car-count tracker belong to the class itself rather than any individual instance, while read-only properties such as a license plate can only be assigned during construction, ensuring immutability afterward. Concrete classes without any hierarchy are appropriate when no meaningful variation in behavior exists across types, as introducing abstract classes or interfaces prematurely leads to unnecessary complexity. A common pitfall is the 'God Class', where a single concrete class accumulates too many responsibilities, violating the Single Responsibility Principle from SOLID design guidelines.

11 Indian-Origin US Citizens Missing During Kailash Yatra After Nepal Floods

A group of 11 US citizens of Indian origin has gone missing while on a pilgrimage to Kailash. The group was being led by Anil Goyal, a US citizen originally from Delhi. The pilgrims were last seen at the Ruswa Garhi immigration centre in China before they disappeared. Their disappearance is believed to be connected to flooding that struck the Nepal region. Authorities are currently working to determine the whereabouts and safety of the missing individuals.

‘Fortunately, my bodyguard retaliated’: Casino director says shooters had recceed his movements

Anthropic permanently increases Claude's weekly usage limits by 25%

Anthropic has announced a permanent 25% increase to Claude's weekly usage limits. The change applies to the AI assistant's usage caps, giving users more access on a recurring weekly basis. The announcement was made via Anthropic's official Bluesky account. The update appears to be a response to user demand for greater access to the Claude AI platform.

SSR Cache Isolation Flaw in Popular React Data Fetching Library Leaks User Data

A security researcher auditing a widely used React data fetching library discovered a cache isolation vulnerability affecting server-side rendering environments. The library initializes its cache as a module-level variable — a single shared Map object created once and reused across all incoming server requests in Node.js. Because modules are loaded once and shared globally, concurrent users hitting the server could inadvertently receive another user's cached data under the same cache key. The flaw requires no malicious input; two legitimate users making simultaneous requests is enough to trigger cross-user data exposure. The researcher found no built-in mechanism such as AsyncLocalStorage or per-request cache factories to enforce request-scoped isolation by default.

How Developers Can Automate AI-Powered Upsells Inside Customer Support Chats

A technical guide published on DEV Community outlines how to build an automated upsell system that triggers personalized product offers during live customer support interactions. The system connects Intercom or Zendesk as the support front-end with OpenAI's language model to generate contextual product suggestions in real time. The workflow is orchestrated using n8n, an open-source automation tool, which routes conversation data to OpenAI, then fetches relevant products from a Shopify store and sends the customer a checkout link. Conversion events are subsequently logged into a CRM such as HubSpot or Salesforce, requiring no manual input from the support agent. Developers familiar with REST APIs are estimated to need six to eight hours to complete the full build.

India Sends Gujarat Forensic Team to Nepal to Identify Flood Victims

A forensic expert team from Gujarat has been dispatched to Nepal to help identify victims of recent flooding. The mission is being carried out under the joint oversight of India's Ministry of External Affairs and Ministry of Home Affairs. The team will apply Disaster Victim Identification protocols directly at the affected sites in Nepal. The deployment reflects India's cross-border humanitarian and disaster response cooperation with its neighboring country.

Tea Break and Visa Delay Save 28 Indian Pilgrims From Nepal Flash Floods

Twenty-eight pilgrims from three Indian states narrowly escaped deadly flash floods in Nepal and have since returned home safely. A short tea break and a resulting ten-minute delay kept one group away from the flood-affected area at a critical moment. A separate group was inadvertently protected when a visa delay prevented them from entering the flood zone. The pilgrims witnessed the floodwaters sweeping away buses and entire villages. Authorities are now working to rescue additional pilgrims from Andhra Pradesh who remain stranded in the region.

How one dev team cut redundant CI runs in a merge queue to speed up delivery

A software development team found that their merge queue, adopted to speed up code delivery, was running a full test suite separately for each pull request waiting to land — meaning four queued PRs triggered four complete runs, even though the last already included the others. Their test suite comprised over 15,000 tests taking 13 to 19 minutes per run, with 93% of CI time consumed by the test step alone. The team discovered that the slowest 1% of tests — around 226 that spin up real processes and migrate databases — accounted for 41% of total test time, making simple test deletion ineffective. After moving compute to idle owned hardware, the same suite dropped from up to 16 minutes to 108 seconds. The article examines whether redundant runs in a merge queue can be safely eliminated and what approaches to stopping them are viable.

John Cena admits he and The Rock genuinely hated each other before WrestleMania 28

WWE legend John Cena has openly revealed that he and Dwayne 'The Rock' Johnson harbored real animosity ahead of their WrestleMania 28 bout. Their rivalry extended beyond scripted storylines, with personal insults exchanged and tensions reportedly boiling over backstage. Despite publicly projecting confidence, Cena privately acknowledged that The Rock was likely to win the match. Following the contest, both men exchanged apologies that laid the groundwork for mutual respect. The two have since gone on to be regarded as icons of the modern WWE era.