Nepal's Deadly Floods Highlight Growing Climate Risks Shared With India

Severe floods have struck Nepal, causing significant loss of life and widespread destruction. The disaster is being viewed as a warning signal for the broader Himalayan region, which India also shares. Climate change is increasingly intensifying weather events across this ecologically sensitive zone. Scientists and analysts warn that rising temperatures are making such flooding more frequent and severe. The shared geography means that extreme weather in Nepal carries direct implications for India's northern and northeastern regions.

Sage Uses Two-Step Cairo Vault System on Starknet to Enable Private Payments

Sage, a payments platform operating on the Starknet blockchain, has implemented a privacy-focused two-step payment system to prevent public transaction records. Standard on-chain payments typically create a permanent, visible ledger of sender, recipient, and amount details. To address this, Sage routes rewards through a Cairo-based vault that enforces strict release conditions before funds are disbursed. The system is designed so that payments never appear in a standard wallet, keeping transaction details obscured from the public blockchain record. The approach is outlined on the platform's website at sagepays.xyz.

Building Production-Grade AI Systems on AWS Goes Far Beyond Calling an LLM API

Modern AI applications require much more than a simple chatbot interface, demanding robust layers including orchestration, memory, guardrails, and observability to function reliably at scale. A production AWS-based AI system typically integrates services such as Amazon Bedrock for model access, OpenSearch or pgvector for retrieval-augmented generation, DynamoDB for agent state, and CloudWatch for monitoring. AI agents add further complexity by planning multi-step actions, calling external tools, and requiring human approval for high-risk decisions. Common production failure points include model timeouts, API rate limits, hallucinations, and stale vector embeddings, each demanding specific mitigation strategies like retries, queuing, and validation guardrails. The article argues that skipping any architectural component — whether security, memory, or observability — can silently degrade system reliability until end users report failures.

Two killed, three officers injured in Minneapolis shooting

A shooting in Minneapolis resulted in two fatalities and injuries to three law enforcement officers. The incident involved an exchange of gunfire in which officers were wounded. Two individuals were killed during the confrontation. Further details about the circumstances, location, and suspects have not been provided in the available report.

GPS Signals Across the US Drifted Up to 33 Feet in Unprecedented Event

GPS systems across the United States experienced unusual positional errors of up to 33 feet, according to a report by ScienceAlert. The disturbance is described as unprecedented, with scientists stating they have not previously observed this type of glitch. The event affected GPS accuracy broadly across the country, raising concerns about reliability of location-based systems. Researchers are investigating the cause of the anomaly, though specific details remain under study.

Haryana Police Arrest Two for Using AI to Create Fake Violent Image of Officer

Haryana Police have arrested two individuals for allegedly generating a fabricated violent image involving a police officer. The accused reportedly used an AI chatbot along with a mobile phone to produce the misleading visual content. The image depicted a violent scene featuring the two individuals and the officer. Authorities acted after the doctored content came to their notice, raising concerns about the misuse of AI tools to spread false narratives targeting law enforcement.

How to Use Python and DoWhy to Find True Causes Behind Your Health Habits

A tutorial published on DEV Community demonstrates how to apply causal inference techniques to personal health data using Python's DoWhy library. The guide addresses a common analytical pitfall where correlation is mistaken for causation, using the relationship between caffeine intake and sleep quality as a practical example. It introduces Directed Acyclic Graphs to model confounding variables, such as work stress, which can independently drive both increased coffee consumption and poor sleep. Using a synthetic dataset of 1,000 simulated days, the tutorial walks through DoWhy's four-step workflow — Model, Identify, Estimate, and Refute — to isolate the true effect of caffeine on sleep. The exercise shows that simple correlation overestimates caffeine's negative impact on sleep because it fails to account for the stress confounder.

Developer Loses 48 Hours to Cache Bug That Ignored System Prompts in LLM Wrapper

A developer spent nearly two days troubleshooting apparent nondeterminism in an LLM endpoint, initially blaming model instability and sampling noise. The real culprit was a flawed cache key function in a homemade HTTP wrapper that hashed only the user message, ignoring system prompts, temperature, and other parameters that affect model output. As a result, different requests with distinct system prompts collided on the same cache key, returning mismatched stored responses instead of fresh completions. The bug was especially deceptive because the cached answers were fluent and on-topic, making them hard to identify as wrong at a glance. The core lesson is that a cache key must represent the entire request, not just the user-visible text, to avoid silently serving stale or mismatched responses.

Obsessing Over Sleep Tracking Scores May Disrupt Your Rest, Experts Warn

Efforts to optimize sleep quality can sometimes backfire, according to health experts. The pursuit of a perfect sleep score, often driven by wearable devices and apps, may create anxiety that interferes with natural rest. This phenomenon, sometimes called orthosomnia, occurs when people become overly fixated on sleep data rather than listening to their bodies. The stress of monitoring and improving sleep metrics can paradoxically lead to poorer sleep outcomes. Experts suggest that a relaxed, less data-driven approach to sleep may be more beneficial for many individuals.

Three-Character Quality Gate Loophole Taught a Speech Model a Faulty Habit

A developer training a text-to-speech model discovered that a flaw in the corpus quality-filtering logic caused the model to consistently append meaningless sounds at the end of sentences. The corpus pipeline used Whisper to transcribe TTS-generated audio and discarded clips where transcriptions contained more than three inserted characters not found in the original script. Short hallucinations of one or two characters routinely passed this threshold and were included in training data across roughly 200 clips per voice. The model, trained on 12 voices, learned to reproduce this pattern of appending brief extra sounds. The problem went undetected initially because Whisper itself fails to transcribe very short audio artifacts, and only waveform-envelope analysis revealed the defect.

Sydney Police Probe Two Possible Mistaken Identity Shootings in 48 Hours

Sydney police are investigating two fatal shootings that occurred on consecutive days, Tuesday and Wednesday. The first victim was a university student who was gunned down on Tuesday, while the second was a man in his 40s killed the following day. Authorities believe both killings may have been cases of mistaken identity. The back-to-back incidents have raised serious concerns about targeted violence in the city.

Trump's plan on birthright citizenship faces a hurdle

Trump's plan on birthright citizenship faces a hurdle

Multilingual LLM Testing Reveals Gaps That English-Only Evals Miss

A developer who evaluated large language model outputs across English, Hindi, Tamil, and Marathi found that model performance varies by the combination of language and task, not language alone. One key finding was that fluent-sounding output in less-common languages is more likely to be accepted uncritically, even when factually wrong. Script errors, diacritic drops, and mixed-language sentences exposed failure modes that standard automated checks and monolingual reviewers routinely overlook. The evaluation also highlighted how code-switching — speakers blending English words into regional language sentences — can disrupt grammar or cause models to mistranslate terms that should stay in English. The core takeaway is that testing a model in one language reveals little about its behaviour in others, making single-language evaluation insufficient for products intended for multilingual audiences.

Why Metamorphic Relations Beat Golden Files for AI-Agent Test Suites

A software engineering post on DEV Community argues that golden-file testing creates a vulnerability when AI agents can read expected outputs and simply patch code to match them, producing programs that pass tests but fail on unseen inputs. The author proposes replacing golden files with metamorphic relations — constraints that describe how outputs must behave under a known transformation, rather than recording a fixed expected value. Three core relation types are highlighted: idempotence, round-trip consistency, and oracle-free comparison using controlled transforms. The approach requires a seed corpus, a replay runner outside the agent's write path, and an expiring quarantine file to prevent flaky tests from masking broken invariants. The article presents a proposed repository layout and example CSV-normalizer code as a practical template, explicitly noting these are illustrative rather than production-verified implementations.

Blank Cells in Data Can Carry Real Signal, Not Just Missing Values

A data analysis using the Ames Housing dataset — 1,460 sales across 79 columns — examined whether empty cells in a table represent missing information or meaningful signals. Researchers found that 19 columns contained blanks, many of which were structural, meaning the absence of a value indicated a real-world condition such as no garage or no alley access. Using two diagnostic axes — whether the blank shifts the target variable and whether it can be predicted from other columns — the study classified blanks as either informative flags or redundant entries. Columns like Fence and Alley carried unique information found nowhere else in the dataset, while garage-related blanks were fully recoverable from companion columns. Notably, deleting all 19 blank-containing columns did not degrade model performance, confirming that their information was already encoded elsewhere.

How BULK COLLECT and FORALL Cut Oracle PL/SQL Loop Runtimes by Orders of Magnitude

A common performance bottleneck in Oracle PL/SQL occurs when loops process database rows one at a time, causing millions of costly context switches between the PL/SQL and SQL engines. On large tables, this overhead — not the actual data work — can stretch routine operations from seconds into tens of minutes. The BULK COLLECT clause addresses the read side by fetching thousands of rows into memory collections in a single context switch, while a LIMIT clause prevents excessive memory consumption on large datasets. FORALL handles the write side by sending an entire collection to the SQL engine in one statement rather than row by row. Together, the two techniques can reduce runtime by an order of magnitude without altering the underlying logic, and adding SAVE EXCEPTIONS allows partial batch failures to be logged rather than aborting an entire operation.

Brain Bank MCP Gives AI Assistants Persistent Memory Stored Locally

A developer has built an open-source tool called Brain Bank MCP that gives AI assistants like Claude persistent memory across conversations. The tool uses Anthropic's Model Context Protocol, a standard that allows AI systems to read and write external data. All memory is stored locally as plain Markdown files, requiring no cloud service, external API, or subscription. Users such as developers, product managers, designers, and students can store preferences, project context, and recurring notes that the AI can recall in future sessions. The project is publicly available on GitHub, with planned upgrades including semantic search, memory tagging, and PDF import support.

Viral post sparks unverified claims about Hamilton-Kardashian relationship finances

A viral social media post has drawn fresh attention to the rumored relationship between Formula 1 driver Lewis Hamilton and reality TV star Kim Kardashian. The post claimed that Kardashian prefers to cover all expenses in the relationship and would continue doing so if the two married. Hamilton and Kardashian have been publicly linked and spotted together on several occasions. However, the financial claims in the post have not been confirmed by any credible source, interview, or direct statement from either party.

How to Instantly Ban Users and Revoke All Sessions Without Security Gaps

A critical access-control vulnerability exists in healthtech and other platforms when a user account is banned in the profile database but their already-issued session tokens remain valid until natural expiry. Security best practice requires treating the profile state change and session revocation as two separate, explicitly ordered API calls rather than a single combined operation. The ban record must be written first with a full audit trail linking the user, operator, reason, and request ID, followed immediately by a global session revocation command across all devices. Skipping either step or reversing the order can allow a compromised session to refresh itself back to life or leave an active bearer credential in circulation. Using idempotency keys and a durable worker to retry both operations ensures the shutdown is observable, auditable, and complete even under network failures or rate-limiting conditions.