Skydiving trip ends in tragedy, as 11 die in France plane crash

NEET Leak Row Exposes Telegram's Dual Role in India's Shadow Education Economy

Telegram's temporary ban in India, triggered by its alleged role in NEET exam paper leaks, has reignited debate over the platform's place in the country's education landscape. While many students rely on Telegram for affordable access to study materials, the platform also enables widespread sharing of paid educational content without authorization. This has drawn criticism from copyright holders and sparked legal battles over intellectual property in the education sector. Critics of the ban argue it penalizes legitimate users without tackling the deeper systemic causes of exam leaks. The controversy highlights the difficult challenge policymakers face in balancing intellectual property protections against the need for broader educational access.

11 Dead After Civilian Aircraft Crashes in France

A civilian aircraft crash in France has resulted in 11 fatalities, according to a report by the Economic Times. The incident involved a non-military aircraft, though further details about the type of plane have not been specified in available reports. The exact location within France and the cause of the crash have not been confirmed in the provided information. Authorities are expected to investigate the circumstances surrounding the deadly crash.

Drin launches agent-native email infrastructure with full send, receive, and parse support

A developer has built Drin, an email infrastructure platform designed specifically for AI agents that need full two-way email capabilities. Unlike traditional email APIs that focus only on outbound sending, Drin provides agents with a real inbox supporting send, receive, inbound parsing, thread context, and reply workflows. The platform also includes webhook support, a CLI, MCP tool integration, and an approval gate for risky outbound messages. Drin is currently live with a free sandbox tier, while production plans start at $3 per month. The creator is seeking feedback from developers building AI agents, CRMs, support tools, or any application requiring programmable inbound email.

Speculative Decoding Benchmarked on CPU: Acceptance Rates Vary Sharply by Task

A developer ran a controlled benchmark of Speculative Decoding (SD) using Qwen2.5-0.5B as the draft model and Qwen2.5-1.5B as the target, testing across code, JSON, and story generation tasks on a CPU-only machine. SD was 49–62% slower than standard autoregressive generation across all task types, consistent with the theoretical inequality that governs when SD wins or loses. Mean token acceptance lengths differed significantly by task: JSON scored highest at 3.50, code at 3.00, and creative story generation lowest at 2.11, reflecting how structured tasks are easier for draft models to predict. A key finding was that 15–30% of draft rounds resulted in zero accepted tokens, meaning the system paid full compute cost for both draft and target passes while producing only a single token. The author notes that while CPU speed numbers are not directly transferable, the acceptance-length patterns are relevant to GPU deployments and suggest task type is a stronger predictor of SD gains than model size alone.

Why AI Support Bots Without Memory Are Just Costly FAQ Pages

A growing critique in the AI development community argues that support bots lacking conversational memory fail to deliver genuine assistance. Without the ability to retain context from past interactions, these bots cannot build on previous exchanges or personalize responses. Critics contend this makes them little more than expensive, automated FAQ pages rather than true support agents. The discussion highlights a key limitation in current AI agent design, particularly around memory and continuity. Developers are increasingly calling for better memory integration in large language model-based support systems.

Unity developer builds custom A* pathfinder for procedurally generated mazes

Abdul Fahad Shahbaz, a Unity game developer from Lahore, Pakistan, has shared how he implemented a custom A* pathfinding system for a procedurally generated maze game. Because the maze layout is created fresh at runtime each session, Unity's built-in NavMesh system — which requires pre-baked static geometry — was not a viable option. He combined iterative Depth-First Search (DFS) to generate the maze with a custom A* algorithm to navigate it, both working directly on a shared grid data structure. DFS was chosen over alternatives like Prim's algorithm because it produces longer, winding corridors better suited to a time-pressure racing game. Shahbaz published the full C# implementation, including a self-contained AStarPathfinder class, on DEV Community for other developers to reuse in their own Unity projects.

Ram Temple Donation Theft Accused Splurged on Farmhouse and SUV, Probe Finds

A man named Anukalp is accused of orchestrating theft of donations at the Ram Temple in Ayodhya. He allegedly arranged for his brother-in-law Lavkush to be placed on the team responsible for counting cash donations from devotees. The accused reportedly used the stolen funds to finance a noticeably lavish lifestyle. Investigators found that the ill-gotten money was spent on assets including a farmhouse and a Scorpio SUV. The case has raised serious concerns about the security and oversight of donation collection at the prominent religious site.

AEGIS: AI Crisis Command System Built With Persistent Memory for Disaster Response

A team participating in the Agent Memory Hackathon has developed AEGIS (Artificial Earth Global Intelligence System), an AI-powered tactical command platform designed to address memory limitations in global crisis response. The system combines real-time 3D geospatial visualization with a persistent semantic memory engine, allowing it to retain and recall past operational decisions such as failed evacuation routes or compromised supply hubs. AEGIS uses Groq for fast reasoning and Vectorize Hindsight to store strategic knowledge, enabling operators to issue natural-language commands that trigger parallel actions on a live 3D globe. The project targets what its creators call 'institutional amnesia' — the loss of critical operational knowledge that typically resides in exhausted personnel, outdated documents, or unread postmortems. The team positions AEGIS as a prototype for India's first AI-driven crisis command infrastructure capable of learning from past disasters rather than starting each emergency from scratch.

Java Engineer Builds Python Countdown Timer CLI to Master Time Control and Exception Handling

A Java developer documenting their journey learning Python from scratch has built a countdown timer command-line tool as the third project in their Python series. The tool lets users choose a countdown duration from a menu, ticking down one second at a time while overwriting the same terminal line using carriage return characters. The project focused on key Python concepts including time.sleep(), reverse loops with range(), and exception handling using try/except/finally blocks to manage interruptions like Ctrl+C. The developer also explored testing techniques such as mocking time.sleep to deliberately trigger exceptions and capturing standard output with pytest's capsys fixture. The article is part of an ongoing series where the author shares honest reflections on their learning process, clearly distinguishing between self-written code and AI-assisted guidance.

n8n Notion Node Lets You Automate Database Tasks Without Code or Zapier Fees

The n8n Notion node enables users to automate reading, creating, updating, and searching Notion database pages without manual effort or paid third-party tools. To get started, users must create a Notion integration token, connect it to their database, and add the credentials inside n8n. The node supports two resource types — Database Page and Block — with operations including Get Many, Get, Create, and Update. Filters, compound queries, sorting, and automatic pagination are all supported, allowing precise control over which database records are retrieved or modified. A free workflow JSON template is provided to help users implement the setup immediately.

India rejects Pakistan's bid to link it to Karachi militant attack

India has strongly denied Pakistan's allegations connecting it to a recent terrorist attack in Karachi. The Indian Ministry of External Affairs called on Pakistan to dismantle its own terror infrastructure instead of making unfounded accusations. The attack targeted a paramilitary headquarters and resulted in casualties. Pakistan has threatened retaliation and has previously carried out strikes in Afghanistan, which Kabul has denied. New Delhi also criticized Islamabad for allegedly using terrorism as an instrument of state policy.

One million passports leaked online

https://web.archive.org/web/20260617075233/https://www.theve... https://archive.ph/n2HnK Comments URL: https://news.ycombinator.com/item?id=48706389 Points: 182 # Comments: 106

How Component Separation Makes Screenshot Tests and Benchmarks Stable in KMP

A software architecture post on DEV Community explains why screenshot tests and on-device benchmarks are the most unreliable tiers in mobile test suites, often failing due to non-determinism in the components under test rather than faults in testing tools like Paparazzi or Macrobenchmark. The author argues that instability stems from UI components directly depending on live data sources, real clocks, animations, and async loading, all of which introduce variable inputs that produce inconsistent outputs. The proposed fix is to treat every screen as a pure function of its state, ensuring composables accept only a frozen UiState and never reach for data internally. By routing all side-effectful dependencies such as network calls, location, and clock reads through abstracted ports in a shared CoreLib, tests can inject fully controlled, deterministic inputs. This approach guarantees that a golden screenshot changes only when the UI genuinely changes, and that benchmark numbers are not skewed by I/O jitter or warm-up variance.

Insider Claims Taylor Swift and Travis Kelce May Have Secretly Married

An insider has claimed that Taylor Swift and Travis Kelce may have already exchanged vows, sparking fresh wedding speculation. The couple is also rumored to be planning a ceremony at Madison Square Garden on July 3. Neither Swift nor Kelce has confirmed any of these reports. The alleged wedding timeline is said to be arranged around Kelce's NFL offseason schedule. According to reports, the couple plans to delay an extended honeymoon to allow Kelce to prepare for the Kansas City Chiefs' upcoming season.

Qwen-AgentWorld Turns a Language Model into a Fast RL Training Simulator

Researchers released Qwen-AgentWorld on June 24, 2026, introducing a language model trained to function as a world model for reinforcement-learning agents. Given a current observation and an action, the model predicts the next environment state, effectively replacing the need for a live environment during training. This decoupled approach allows thousands of simulation rollouts to run simultaneously, overcoming the slowness and cost of real-environment RL training. The system also serves as a foundation model, giving downstream agents a warm-start before task-specific fine-tuning. A hybrid reward signal is used in the final RL stage to improve how faithfully the model's predictions match real-world outcomes.

Bengal TMC MLA Humayun Kabir Faces Police Case Over Alleged Threats Against BJP

A police case has been registered against West Bengal TMC MLA Humayun Kabir over alleged threats directed at BJP leaders. The case follows a video that went viral in which Kabir is heard making remarks against local BJP figures. In the clip, Kabir accuses BJP leaders, including Anamika Ghosh, of behaving with the arrogance of elected representatives despite losing the election. The incident has drawn political attention in the state amid ongoing tensions between the TMC and BJP.

Developer Tests Zyloo, a Unified OpenAI-Compatible Gateway for Multiple AI Models

A developer has been evaluating Zyloo, a gateway service that provides a single OpenAI-compatible API endpoint to access multiple AI model providers. The core appeal is eliminating the need to manage separate API keys and integrations for each provider, streamlining model comparison within existing workflows. The reviewer found it particularly useful for development tasks such as coding agents, automation scripts, and bots that already support OpenAI-compatible endpoints. While the concept is considered practical for experimentation, the author advises caution before adopting it in production environments, citing the need to verify latency, reliability, privacy, and pricing. The author disclosed that Zyloo may credit their account if the post is approved through the platform's creator rewards program.

PCB Triples Participation Fee, Introduces Mandatory Contracts in Department Cricket

The Pakistan Cricket Board (PCB) has announced a major overhaul of its departmental cricket structure. The participation fee for the President's Trophy and President's Cup has nearly tripled for the current season. As part of the reforms, player contracts have been made mandatory for participating departments. The changes signal a significant shift in how departmental cricket is administered and funded in Pakistan.

Tech Firms Adopt 'Cut and Pivot' Layoff Strategy to Rebuild Workforces for AI Era

A new layoff pattern dubbed 'Cut and Pivot' is emerging across the tech sector in 2026, where companies simultaneously let go of large numbers of workers in legacy roles while opening far fewer positions requiring entirely different skill sets. Unlike earlier waves of tech layoffs driven by cost-cutting or over-hiring corrections, this strategy reflects a deliberate effort to reshape talent pools toward areas such as AI integration, quantum computing, and next-generation product development. Industry observers note that roles tied to maintaining older monolithic systems are being eliminated to fund smaller, agile teams focused on AI-driven solutions. One cited example involves a bank that reduced a 15-person legacy platform team to four employees while using the same budget to build a new six-person AI/ML unit. Analysts attribute the trend largely to rapid advances in AI automating routine development and infrastructure tasks, combined with investor pressure on companies to demonstrate a credible vision for future growth.