Manipur Naga body urges PM to order judicial probe into alleged killing of six civilians

Why AI Agent Access Without a Rights Framework Is a Hidden Risk

A technical analysis published on DEV Community argues that the true measure of an AI agent's capability is not how many tools it can access, but what actions it is actually permitted to execute. The piece distinguishes between visibility — what an agent can read — and mutation — what it can change — warning that conflating the two creates ungoverned systems. The author proposes a five-layer rights framework covering visibility, mutation, proof of action, escalation conditions, and post-failure permission revocation. Most teams building agentic systems, the article contends, can answer only the first one or two layers, leaving critical gaps in oversight and accountability. The core argument is that authority granted to an agent should be proportional to the consequence of its actions, not simply to the number of integrations it holds.

Developer finds Polymarket bots lose money when tracking Binance instead of Chainlink

A developer running an open-source Polymarket trading bot discovered that using Binance spot prices as a reference was causing real financial losses. Polymarket settles its crypto markets using Chainlink Data Streams and Chainlink Automation, not exchange feeds, meaning the two prices can diverge by 0.3% to 0.8% during volatile periods. In binary markets, even a $2 price difference at the moment of resolution determines whether every open position wins or loses. The developer updated the bot to pull prices directly from Chainlink's Data Streams API and added checks for feed staleness. An additional complication is that Chainlink Automation introduces a small execution delay beyond the listed market end time, especially when the Polygon network is congested.

Developer Builds Quad-Brain Edge AI Framework to Handle Noisy Industrial Sensor Data

A software developer has designed QuadBrain-Nexus, an open-source multi-sensor data fusion framework built for Edge AI hardware such as NVIDIA Jetson devices. The system addresses a common failure in industrial environments where static threshold-based anomaly detection breaks down under high-noise, high-variance conditions. It uses a four-engine architecture — covering frequency analysis, spatial tracking, data ingestion, and Bayesian decision-making — with each component running concurrently on isolated CPU or GPU cores. The implementation relies on Python's multiprocessing module to bypass the Global Interpreter Lock, enabling deterministic sub-millisecond execution loops with vectorized NumPy operations. The framework is designed to be sensor-agnostic, accepting telemetry from diverse hardware including flow meters, industrial sonars, and pressure sensors via UDP or WebSockets.

6.2-magnitude earthquake jolts Afghanistan, tremors felt in Pakistan, Jammu and Kashmir and Delhi-NCR

6.2 magnitude earthquake hits Afghanistan; tremors felt across north India

Developer Builds Open-Source Feature Flag Audit Tool Inspired by Knight Capital's $440M Loss

A software engineer built Tombstone, a self-hosted feature flag intelligence platform, after a 2:47am incident revealed his team had no reliable way to track which flags had changed across 12 services during a payments outage. The core problem was that tools like LaunchDarkly, Jira, and Notion each held partial information but shared no causal model, making it impossible to link a flag change to a downstream latency spike. The project draws on the 2012 Knight Capital disaster, where reactivating a dormant flag key on a single server triggered $440 million in losses in 45 minutes because no system tracked key provenance or blocked reuse. Tombstone is designed to permanently retire, or 'tombstone,' flag keys after deprecation so they cannot be reactivated with outdated semantics in future deployments. The tool targets engineering teams that have outgrown informal audit trails like Slack channels and shared docs but need durable, queryable flag history without relying on fragmented third-party platforms.

Karnataka man dies after detonating crude bombs in car during relationship dispute

A 30-year-old man named Nagendra died in a car explosion on a Karnataka highway following an apparent relationship dispute. Police suspect Nagendra attacked his partner, Ramya, before setting off crude bombs inside the vehicle. Ramya and the cab driver were able to escape the burning car and survived the blast. Authorities are currently investigating the origin of the explosives and the sequence of events that led to the incident.

Developer builds ruleset to strip AI writing patterns from AI-generated text

A developer published a post on DEV Community describing a tool called Alchemy, a Markdown-based ruleset designed to remove telltale AI writing patterns from AI-generated content. The tool targets recurring signals such as em-dash overuse, filler phrases like 'in today's fast-paced digital landscape,' and words like 'delve' and 'robust' that research shows spiked sharply in frequency after ChatGPT's release. Its design draws on empirical sources, including a 2025 study by Kobak et al. finding 'delve' appeared at roughly 28 times its pre-ChatGPT rate in biomedical papers, and Pangram Labs data on inflated phrase frequencies. Alchemy works by flagging density of co-occurring tics rather than penalising any single word, and integrates with AI coding agents via a single init command. The tool is MIT-licensed and available on GitHub for anyone to read or adapt.

Teen developer runs Hack Club software jam to reward genuine creative effort

Fox, a 16-year-old developer, was invited by Hack Club to organize a software event after the organization noticed his online projects. He felt that while Hack Club is good at funding projects, it falls short in recognizing hard work and craftsmanship. In response, he created Radish Jam, a software jam designed to foster meaningful competition and reward well-made work. The initiative drew inspiration from the quality-focused feedback culture found on platforms like Hacker News. Fox documented his organizational process in a blog post, hoping it could serve as a guide for future organizers within or outside Hack Club.

Ireland bowl as West Indies eye semi-final spot

Both teams made a change that alters the team balance

Browser Security Model Explained: SOP, CORS, XSS, CSRF and How They Work Together

Modern web browsers operate as hostile execution environments where untrusted code from one tab can sit dangerously close to sensitive user sessions in another. The browser's security model — built around policies like Same-Origin Policy (SOP) and Cross-Origin Resource Sharing (CORS) — was designed to counter threats such as session hijacking, cross-site scripting, and cross-site request forgery. SOP restricts how scripts from one origin can interact with resources from another, defined by a combination of protocol, host, and port, though it blocks reading responses rather than sending requests, which is why CSRF attacks remain possible. CORS allows servers to deliberately opt in to cross-origin access via HTTP headers, but a common misconfiguration — reflecting any incoming Origin header back without validation alongside credentials — can completely bypass SOP and expose authenticated data. Understanding these security boundaries is considered essential for developers, as attackers routinely exploit gaps in their implementation to steal user data or hijack sessions.

Bharat Taxi to expand to over 500 cities in 2 yrs

Suspicious Discontinuities

Article URL: https://danluu.com/discontinuities/ Comments URL: https://news.ycombinator.com/item?id=48698151 Points: 16 # Comments: 0

Codename One Ships Native Linux App as a Single 5MB Binary for x64 and ARM

Codename One, an open-source framework for building cross-platform apps from Java or Kotlin, has released a native Linux desktop port that compiles to a single self-contained ELF binary. The framework's ParparVM tool translates Java and Kotlin bytecode to C, which is then compiled into a native binary requiring no JVM on the user's machine. The Linux port relies on widely available system libraries including GTK3, Cairo, Pango, GStreamer, and WebKitGTK, covering graphics, media, networking, and browser functionality. To ensure broad compatibility, binaries are compiled against glibc 2.17 from 2013, allowing them to run on virtually any mainstream Linux desktop. The optimized build strips unused code, keeping a typical app around 5MB while launching faster than many native GNOME applications.

Trump could visit India next year…look forward to setting (it) up: Rubio

Apple Raises Prices on MacBooks, iPads and HomePod, Citing AI-Driven Costs

Apple CEO Tim Cook has described the company's current pricing as 'unsustainable' and called further price increases 'unavoidable.' The 16-inch MacBook Pro rose by $300, the 11-inch iPad Air jumped from $599 to $749, and the HomePod Mini increased by $30 to $129. Cook attributed the hikes to surging demand from the AI industry, which has driven up the cost of components like RAM. Apple is not alone — Xbox consoles have risen nearly 25 percent in price and Nothing scrapped a phone launch entirely amid the same pressures. The trend reflects broader market stress as AI infrastructure investment continues to strain global hardware supply chains.

India A lead by 170 as Sai Sudharsan retires hurt against Sri Lanka A

India A extended their lead to 170 runs in their ongoing match against Sri Lanka A. Sai Sudharsan, who had been contributing with the bat, was forced to retire hurt during India A's innings. Earlier, Sri Lanka A were bowled out for 330, with Mohammad Nabi playing a key role with the ball. Nabi claimed four wickets to help India A dismiss their opponents and gain a competitive advantage. India A will look to press further ahead as they continue their second innings.

Developer runs modern LLMs on a 2013 GTX 770 with a five-byte firmware patch

A developer has demonstrated that NVIDIA's Kepler-generation GTX 770, officially abandoned by NVIDIA after driver version 470.256.02, can still run modern large language model inference workloads on Linux kernel 7.x. Two core obstacles were overcome: the legacy driver source code failed to compile against kernels 6.15 and above due to removed APIs, and the CUDA initialization routine returned an error caused by a hardcoded version check inside the proprietary libcuda.so library. Community-sourced kernel patches resolved the compilation issues, while a targeted five-byte binary patch to the shared library bypassed the version mismatch that blocked CUDA from initializing. Additional changes to llama.cpp source code and the use of CUDA 10.2 tooling with Clang allowed the project to build and run successfully on the GPU's older sm_30 architecture. Benchmark results showed the GTX 770 delivering roughly 1.8 times faster prompt processing compared to CPU-only inference, with the project framed as both an e-waste reduction effort and a practical exercise in low-level systems engineering.

11 Design Patterns for Building Polling Agents in AI Assistants

Polling agents are background processes in AI assistant architecture that repeatedly monitor a data source and trigger actions when specific conditions are met. Unlike standard chat assistants that respond only to user queries, polling agents proactively watch inboxes, task lists, GitHub issues, or job queues on the user's behalf. A well-structured polling agent handles five core responsibilities: waking on schedule, reading the source, tracking previously seen state, evaluating whether new data matters, and acting exactly once without duplication. Developers are advised to keep the language model focused on semantic interpretation and language generation, while delegating scheduling, state management, retries, and locking to standard backend infrastructure. Durable state records — storing metadata such as poll ID, source reference, condition, run timestamps, and failure count — are essential for reliable production deployments.