SShortSingh.
0
IndiaNDTV ·

Fadnavis Vows Constitutional Approach to Maharashtra Reservation Demands

Maharashtra Chief Minister Devendra Fadnavis addressed the ongoing reservation debate in the state on Tuesday. He stated that his government is open to hearing demands from various communities regarding quotas. However, Fadnavis made clear that any decisions taken would strictly adhere to constitutional provisions and established rules. The remarks come amid continued pressure over the Maratha community's longstanding demand for reservation.

0
WorldBBC World ·

Rescuers Search Collapsed Cali Building for Survivors After Earthquake

Rescue teams are working through the rubble of Torres del Limonar, a building that collapsed in Cali, Colombia, following an earthquake. BBC Mundo reporters visited the site to document the ongoing search and rescue operation. Families of those trapped, including a distressed brother, have gathered near the scene awaiting news of their loved ones. Emergency workers continue their efforts to locate and free any survivors still beneath the debris.

0
ProgrammingDEV Community ·

Venture Capital vs. Bootstrapping: A Framework for Founders to Decide

The decision between raising venture capital and bootstrapping is fundamentally a strategic choice, not merely a financing one, according to an analysis by Ralvan. Venture capital suits startups targeting markets where the first company to scale captures most of the value, but it comes with expectations of outsized returns and eventual liquidity events. Bootstrapping, by contrast, allows founders to retain full ownership and control, with a wider range of acceptable outcomes including modest but sustainable success. Key trade-offs include speed and headcount versus cash constraints, diluted ownership versus retained equity, and shared board control versus full autonomy. The piece also cautions that raising capital does not solve fundamental problems like lack of product-market fit, and that hybrid options such as revenue-based financing exist beyond the binary choice.

0
ProgrammingDEV Community ·

Google Launches Gemini 3.5 Family at I/O 2026 With New Models, Agents and App Features

Google unveiled the Gemini 3.5 model family at Google I/O 2026, introducing Gemini 3.5 Flash as its flagship fast-action model alongside Gemini Omni for multimodal video output. A new personal agent called Spark is designed to provide round-the-clock assistance and will integrate with Google Workspace tools, with a macOS desktop experience also planned. The rollout also includes Daily Brief, a personalized morning digest feature, and enhancements to Google's Antigravity agent framework for developers and enterprise users. Gemini Omni is beginning to roll out to Google AI Plus, Pro, and Ultra subscribers, though full technical details and regional availability have not yet been disclosed. The breadth of the release signals Google's intent to embed Gemini across its product ecosystem rather than position it as a standalone AI application.

0
ProgrammingDEV Community ·

How Celery Registers, Queues, and Executes Background Tasks in Python

Celery is a Python tool that allows applications to offload time-consuming jobs — such as sending emails or processing payments — to background workers instead of blocking the main app. It operates through four components: a producer (the app), a broker (message queue), a worker (task executor), and an optional result backend. When a task is triggered, the app sends a message containing the task name and arguments to the broker, which holds it until a worker is available. The worker looks up the task name in its own registry and executes the corresponding function, then stores the result in the backend if one is configured. Task registration happens at import time via the @app.task decorator, and in larger projects, Celery's autodiscover_tasks() automates this by scanning installed apps for task definitions.

0
Crypto & Web3CoinDesk ·

CFTC Orders Kalshi to Keep Operating Prediction Markets in New York

The Commodity Futures Trading Commission (CFTC) has directed Kalshi to continue offering its prediction markets in New York despite an ongoing legal challenge. New York state filed a lawsuit against Kalshi last month seeking to block the platform from running sports-related prediction markets. The federal regulator's order effectively pushes back against the state's attempt to shut down Kalshi's operations. The development highlights a jurisdictional tension between state authorities and federal oversight of prediction market platforms.

0
ProgrammingHacker News ·

Suzanne: New AI Tool Aims to Streamline Physical Product Design and Manufacturing

Suzanne is a newly launched AI-powered tool designed to assist with the design and manufacturing of physical products. The platform, accessible at suzanne3d.com, targets users involved in product development and fabrication workflows. It appears to leverage artificial intelligence to simplify or accelerate steps in the product creation process. The tool was shared on Hacker News, where it received modest early attention. Further technical details about its specific capabilities remain limited at this stage.

0
ProgrammingDEV Community ·

Grok Bot Launches Multi-Agent Tool to Automate LinkedIn, GitHub, and Email Tasks

Grok Bot is a newly released multi-agent automation platform that runs on a user's computer and handles tasks across apps like LinkedIn, GitHub, and email. The tool allows users to set up individual bots for specific functions, such as posting on social media, closing GitHub issues, and reading or replying to emails. A content workflow can be configured to automatically publish posts across multiple platforms, including personal sites, Dev.to, LinkedIn, and X, from a single input. The platform offers a guided onboarding experience, suggesting actions based on a bot's name to help users get started without prior planning. Grok Bot is currently available with a free trial, and early users report a notably smooth setup process for connecting third-party services.

0
TechnologyThe Verge ·

FCC Filing Hints at Sonos Ace Ultra Headphones Arriving This Fall

An FCC filing by Sonos, Inc. has revealed details about upcoming wireless headphones carrying the model number S52A. One of the submitted documents includes an oval diagram, likely depicting an ear cup, labeled 'Sonos Ace Ultra,' though the name does not appear elsewhere in the filing. The naming follows the same pattern Sonos used when it introduced the Arc Ultra soundbar alongside the original Arc. Sonos CEO Tom Conrad had previously hinted at a successor to the Ace headphones during a third-quarter earnings call. The filing suggests the new headphones could launch sometime this fall.

0
IndiaTimes of India ·

China cracks down on 'flash marriage' scams after men lose hundreds of thousands

China's rapid matchmaking industry, known as 'flash marriage', is under scrutiny following reports of widespread financial fraud targeting men. In one case, a 37-year-old man lost up to 500,000 yuan before uncovering his wife's alleged concealed past. Authorities have launched a formal crackdown in response to rising concerns about deceptive matchmaking practices. The issue is linked to broader social factors, including China's gender imbalance, which has made some men vulnerable to financial exploitation.

0
ProgrammingDEV Community ·

Developer's AI Pentest Agent Captures CTF Flags but Hits Submission Wall

A software developer building an autonomous penetration testing agent called HALO successfully expanded its toolset from 31 to 42 tools, enabling a full web-recon to flag-capture pipeline on the CTF platform VulnBegin. The agent correctly identified and logged strings matching the expected flag format, effectively completing its objective from a technical standpoint. However, when the captured flags were submitted to the platform, they were silently rejected without any error message or crash. A key engineering breakthrough during development was fixing silent process hangs caused by child scanner tools inheriting the MCP server's stdin pipe, resolved with a single line of code. The developer documented both the success and the submission failure as equally valuable engineering lessons, noting that a comprehensive test suite of 361 passing tests allowed safe mid-engagement refactoring.

0
ProgrammingDEV Community ·

How to Build a Testable Multi-Provider Chatbot API Adapter for US and EU Apps

Developers building US/EU chatbot apps face hidden risks when relying on a single OpenAI-compatible API, as provider differences in tool calls, streaming events, and regional routing can cause failures that only surface beyond a basic demo. A thin routing layer with a single internal contract, a Python adapter, and an evaluation set focused on answer quality is recommended over chasing the cheapest provider label. Data residency decisions must be made before any API key or SDK reaches production, and secrets should never be exposed to the browser. The adapter should translate provider-specific behavior at the boundary so the rest of the application remains unaware of which backend handled the request. Testing must go beyond checking final text responses to cover retrieval markers, tool argument validity, stream terminators, and regional fallback behavior.

0
TechnologyArs Technica ·

Russian Drones Destroyed Near Key European Offshore Gas Installation

Russian drones were discovered in the vicinity of a critical European offshore gas facility, prompting authorities to neutralize the threat by destroying them. The incident is part of a broader pattern of drone activity attributed to Russia targeting European energy infrastructure. Prior to this event, multiple drone incursions had been recorded over Romanian territory. A separate drone attack targeting Germany was also attempted but ultimately failed. The incidents have heightened security concerns around vital energy assets across Europe.

0
ProgrammingDEV Community ·

Google Launches Gemini 3.1 Pro Preview Across Consumer, Developer, and Enterprise Tools

Google announced Gemini 3.1 Pro on February 19, 2026, positioning it as a significant upgrade within the Gemini model family with a focus on improved reasoning for complex tasks. The model is currently in preview and is being rolled out simultaneously across consumer, developer, and enterprise channels rather than as a standalone model release. Developers can access it via the Gemini API, Google AI Studio, Gemini CLI, and Android Studio, while enterprises have access through Vertex AI and Gemini Enterprise. Consumer-facing access is available through the Gemini app and NotebookLM. Google cited a company-reported score of 77.1% on the ARC-AGI-2 reasoning benchmark as evidence of the model's capabilities, with general availability expected following the preview phase.

0
ProgrammingHacker News ·

Study Examines Whether Large Language Models Develop Introspective Awareness

A new research paper published on arXiv investigates the concept of emergent introspective awareness in large language models (LLMs). The study explores whether advanced AI systems may develop a form of self-awareness as a byproduct of their scale and training. Researchers are examining how LLMs represent and reason about their own internal states. The findings contribute to ongoing scientific debate about the cognitive and metacognitive capacities of modern AI systems.

0
ProgrammingHacker News ·

AI Tools Now Cracking Cybersecurity CTF Challenges Within Minutes

Artificial intelligence systems are demonstrating the ability to solve Capture The Flag (CTF) cybersecurity challenges in a matter of minutes. CTF competitions are widely used in the security community to test skills in areas such as cryptography, reverse engineering, and exploitation. The rapid problem-solving capability of AI in these contexts highlights growing concerns and opportunities around automated security tools. This development raises questions about the future of cybersecurity training and the potential misuse of AI in offensive security scenarios.

0
ProgrammingDEV Community ·

How Redirect Chains and Fake Domains Can Mislead Users Before They Log In

A redirect chain occurs when a browser passes through one or more intermediate URLs before reaching a final destination, a process that is sometimes legitimate but can also signal deceptive intent. Security-focused guidance published on DEV Community explains that multiple redirects across unrelated domains, unusual URL structures, and a final page that differs from the promised content are all worth closer inspection. The article notes that HTTPS and a padlock icon do not guarantee a site is genuine, since fraudulent domains can also obtain valid TLS certificates. Typosquatting — registering domains with minor spelling variations — is highlighted as a common tactic used in phishing pages that can easily go unnoticed on a quick read. Users are advised to verify the exact domain name, check for misleading subdomains, and trace the full redirect path before entering any sensitive credentials.

0
IndiaNDTV ·

Trump Secretly Switched Aircraft in Turkey to Evade Reported Iranian Threat

US President Donald Trump covertly changed planes while on the ground in Turkey on July 8, according to a Washington Post report published Monday. The move was part of a security operation designed to counter an alleged assassination threat. Trump's security team orchestrated the aircraft switch as a precautionary measure. The report did not disclose further details about the nature or credibility of the threat, which was attributed to Iran.

0
ProgrammingDEV Community ·

How one team built shared memory for a multi-model AI agent fleet

A development team running a fleet of AI agents — each backed by a different language model — found that standard per-agent context windows caused repeated work, as no single agent retained knowledge discovered by another. The core problem was that passing more conversation history failed because a large context window on one model cannot be accessed by agents running on different models. Their solution was a model-agnostic shared memory system built on two layers: an append-only event log as the source of truth, and a derived index that agents actually query. Key design requirements included attribution of every memory entry to a specific agent and session, strict scoping so agents only access relevant context, and full auditability for regulated workflows. The team noted that most retrieval failures were actually scoping failures, and that narrowing the searchable candidate set per agent proved more effective than improving ranking algorithms alone.

0
ProgrammingDEV Community ·

Free Hands-On Lab for Microsoft Copilot Studio AB-620 Certification Launched

A developer has released a free hands-on lab for the AB-620 Copilot Studio certification exam on an independent, open-source platform called Examplar. The lab walks learners through creating an agent, writing instructions, testing prompts, publishing, and cleanup — with checkpoints at each step. Alongside the lab, a public preview section offers 25 original practice questions with no reliance on exam dumps. The project is self-funded and independent, with optional paid content packs also available. The free lab and practice questions are accessible at examplar.app.

← NewerPage 85 of 2480Older →