SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World

Programming

LatestTopDiscussed
0
ProgrammingDEV Community ·

Ahsay Issues Second Round of Fixes for Two Chained Flaws in Backup Server Console

Ahsay has released new hotfixes for two linked vulnerabilities in AhsayCBS, the web console used by managed service providers to run backup servers. The flaws, CVE-2026-105133 (an authentication bypass) and CVE-2026-105134 (a critical command injection), were disclosed on 4 October 2026 and can be chained to gain system-level code execution. Huntress reported attacks against internet-exposed appliances starting 7 October, including webshells and a cryptocurrency miner. Ahsay acknowledged its earlier patches did not fully close the flaws and directed users to hotfixes 10.3.4.45 and 9.15.2.70. CISA added both vulnerabilities to its known exploited catalogue on 14 October.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

A CLIP score of 0.85 does not equal 85% listing accuracy

A high confidence score from an image-text model can seem more authoritative than the evidence behind it. This article examines a marketplace verifier from Eşya Bul that uses CLIP to compare listing photos against short text descriptions and keeps listings whose best positive score exceeds 0.85. The code shows the score only answers which supplied description best matches a photo, not whether a seller is honest, an item works, or payment is safe. Because softmax normalizes scores against the candidate set, the same logit can yield a much lower share when a strong alternative is added, and scores from different product groups are not directly comparable. The author recommends testing narrow, human-applicable labels, such as whether an image shows a complete appliance rather than a replacement part, and analyzing errors separately for category selection and the English-language descriptions.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Student fails OS exam by four marks, then passes cloud certification with 912 score

A student's friend narrowly failed an operating systems exam by four marks, yet in the same week passed a first cloud certification scoring 912. The author reflects on feeling torn between university theory and hands-on cloud work, initially assuming one path was wasteful. They found that textbook concepts like deadlock, where processes hold and wait on resources, explained a real Lambda timeout in production. Similarly, normalization habits from coursework caused inefficient DynamoDB data modeling, which raised read costs and led to expensive scans. The experience convinced the author that academic fundamentals and practical cloud skills complement each other.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

BeyondTrust PAM Appliance Backdoor Survived Reboots in Chinese State-Backed Breach

A China-linked state-backed group spent about seven weeks inside BeyondTrust's privileged access management appliance before detection. Google's Threat Intelligence Group and Mandiant say the flaw, tracked as CVE-2026-102764, allowed unauthenticated file uploads that led to root access and theft of the credential vault. The attackers planted a boot-time script that re-adds a passwordless root entry to the sudoers template, so the backdoor rebuilt itself after reboots and manual edits. Mandiant advises rebuilding suspected appliances from a known-good image rather than attempting cleanup. Administrators are urged to confirm they run patch BT26-05 or later, especially self-hosted appliances without automatic updates.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Connecting CrewAI Agents to Alfresco Documents via OpenDMA and LangChain Tools

This guide explains how to give a CrewAI agent access to documents stored in Alfresco without ingesting them into a local vector store. The approach routes the agent through LangChain tools that call an OpenDMA REST API, which in turn talks to an Alfresco adaptor. The toolkit is configured once with the endpoint and credentials, so the agent never handles them as tool inputs. It exposes operations such as listing sites, searching, listing folder contents, reading metadata and reading transformed text. The example task asks the agent to identify the top three risks across a sample Project Apollo site and cite supporting documents. A runnable sample project is available in a companion repository.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How to Collect Coles and Woolworths Grocery Prices and Specials Without Coding

Australian grocery prices are widely debated, but neither Coles nor Woolworths offers a public API for product data. This guide explains how to pull prices, specials, unit prices and barcodes from both chains into spreadsheets or code using third-party scraping tools, without writing a custom scraper. The author discloses that they built the two tools featured and publish them on the Apify Store, and notes that both offer free trials. The walkthrough covers keyword searches, exporting results to Excel or CSV, and scheduling weekly runs to track specials, with an API option for developers.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

Fizzpoker Launches Browser Poker With Friends, No Login Required

Fizzpoker is a new online poker platform that lets users play with friends without creating an account. It was shared on Hacker News, where it gained five points and drew no comments at the time of posting. The source material provides only the site link and basic engagement metrics, so details about its game features, business model, or technology are not available. The service appears aimed at casual, low-friction play among acquaintances rather than public competitive poker.

0 comments Read more at Hacker News
0
ProgrammingHacker News ·

No official XScreenSaver GitHub repository exists, per Hacker News discussion

A Hacker News post highlights the claim that there is no official GitHub repository for XScreenSaver, the long-running screensaver utility for Unix-like systems. The linked GitHub repository under the xscreensaver-official name is referenced as the item being discussed. The submission has gathered only a few points and no comments at the time of reporting. The article text provided contains no further details about who maintains the project or why the repository status is in question. Readers would need to consult the linked pages for any additional context.

0 comments Read more at Hacker News
0
ProgrammingHacker News ·

Mounting Setbacks Raise Questions for Saudi Crown Prince Muhammad bin Salman

A recent Economist article argues that a series of difficulties is currently accumulating for Saudi Arabia's Crown Prince Muhammad bin Salman. The piece, published under the Middle East and Africa section, frames these problems as a broad pattern rather than isolated events. The source material provided here is limited to the article's link and Hacker News discussion metrics, so the specific setbacks cannot be detailed from this information alone. Readers seeking the particulars should consult the full Economist report.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

Engineer Questions How Automation Gains Weigh Against Job Losses and Economic Security

A software engineer writing on DEV Community reflects on the tension between automation's efficiency benefits and its societal costs. The author argues that economic incentives reward replacing human workers, which can lead to job displacement, rising inequality, and social strain. The piece describes this as an "automation paradox," where tools meant to free people from labor still leave survival tied to work. The author suggests that existing systems have not kept pace with technological change and questions whether the goal is broader human benefit or simply individual advantage. The article offers reflection rather than a definitive solution.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Former Aviation Professional Turns to Software Development at 30 With Structured Learning Plan

A detailed guide examines the challenges and strategies for switching from aviation to software development at age 30. It argues that a structured progression from foundational to advanced topics helps build a solid knowledge base. The piece recommends selecting high-quality learning materials and combining practice exercises with projects to bridge theory and application. It also highlights how existing skills such as problem-solving, attention to detail, and procedural discipline can ease the transition, while noting time, financial, and stress-related constraints that need careful planning.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

AsyncShield Enterprise Offers Executives an Air-Gapped Tool to Triage Async Messages

A submission to the Hacktoberfest Open-Source AI Challenge describes AsyncShield Enterprise, a confidential triage tool for engineering leaders facing overwhelming asynchronous communication. The author argues that incident channels, automated alerts, and long email threads bury leaders in messages that rarely require their input. The tool condenses chat transcripts into a two-line brief, generates canned replies, and ranks urgency using Prior Labs' TabPFN-3.5 model. It also offers an audio morning briefing and runs as an air-gapped, zero-telemetry single-page web app to protect proprietary data.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Why AI Agents That Pass Demos May Fail in Production Workflows

An AI agent can impress in a prototype by reasoning through tasks and calling tools, yet real business deployments raise harder questions. Failures such as API timeouts, malformed tool data, duplicated retries, invalid parameters, and required human approvals all challenge reliability. The article argues that making an agent's actions dependable matters more than making the model capable, since a model's proposed action or reported success does not confirm validity or completion. Relying on prompt instructions alone is insufficient, because prompts provide no transactional guarantees, access control, or durable execution records. Instead, production systems need explicit execution semantics, enforced constraints, durable state, and verifiable records of what actually occurred.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Shopify Rollouts Lets Merchants Test Discount Offers Before Black Friday

Shopify announced on October 2, 2026 that discounts can now be included in Rollouts, letting merchants show a new promotion to part of their store traffic. The feature supports comparing a new offer against the existing one, which can help merchants judge its value before a storewide sale. The article notes that higher order counts after a price cut do not automatically mean a more profitable promotion. Rollouts offers three types: Experiment for comparison, Event for time-limited changes, and Launch for permanent ones. Access depends on plan level and staff permissions, with Experiments requiring a higher tier than basic Rollouts.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Solo developer's OpenRouter-style gateway for Chinese open-weight LLMs draws 500 signups but few paying users

A developer built A2Agent, a pay-as-you-go LLM API gateway similar to OpenRouter but centered on Chinese open-weight models such as DeepSeek, GLM, Kimi, Qwen, and MiniMax. The platform has attracted about 500 registered users, yet very few have made paid usage. The developer suspects the audience may be mismatched, since many users seem to want models like GPT and Claude, which the service does not offer. The post invites candid feedback from the community on the product's market fit and the reasons for low conversion.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Claude Code needs a language server to read code accurately, not just text search

An author found that the line numbers in their task notes were all out of date because the code had shifted since they were written. Claude Code's default search works like Ctrl+F, so it can return false matches such as similarly named functions or comments, leading to confident but wrong answers. A language server, the tool behind editor features like go-to-definition and find-all-references, understands code structure and can locate real callers and definitions by name. The post explains how to install a language server such as Pyright or TypeScript's server and connect it to Claude Code via a plugin. It also suggests adding an instruction to CLAUDE.md to find code by name rather than line number.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Deploying a Claude-Powered Coding Assistant on AWS with CDK Pipelines

Deploying AI coding assistants typically involves manual work across Docker builds, secret handling, and ad hoc scripts. This guide shows how AWS Cloud Development Kit (CDK) Pipelines can treat the whole Claude-based agent as infrastructure code, automating build, test, and deployment. The approach relies on a pipeline that pulls source code, builds artifacts such as Docker images, runs tests, and deploys to the target environment. Managing the pipeline as a single, version-controlled source of truth is presented as a way to make deployments repeatable and auditable. The walkthrough includes setting up a small TypeScript CDK project and a command-line script that calls Claude, packaged in a container.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

SlowMist Says Qubic Wallet Network Headers Exposed Private Key Seeds

Security firm SlowMist reported on October 20 that a flaw in the Qubic Wallet could let attackers recover user private keys from network traffic after observing as few as six packet headers. The weakness stemmed from the wallet relying on JavaScript's Math.random(), a non-cryptographic generator, for both account seed creation and network header values. Those header values were sent over unencrypted TCP connections, allowing researchers to reconstruct the generator's internal state in about 15 milliseconds. The findings came after Qubic disclosed on October 18 that roughly 2.94 trillion $QU had been drained from 1,404 wallets through weak randomness in key generation. Qubic had previously paused its network after the unauthorized transfers were first detected.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Total.js Forge: un recetario de código donde cada receta se ejecuta antes de publicarse

El equipo de Total.js ha lanzado Total.js Forge, un recetario para Total.js 5 y jComponent 20 creado para evitar que la documentación quede desfasada respecto al framework. Cada receta es una página Markdown con código, rutas de archivos y un bloque de prueba oculto, y se valida automáticamente creando una aplicación temporal y ejecutando comprobaciones; si una falla, no se publica. Total.js 5 es un framework de Node.js sin dependencias externas que incluye de serie rutas, WebSocket, autenticación, base de datos NoSQL embebida y otras funciones. El sitio ofrece más de cien recetas de Total.js 5, más de treinta de jComponent y páginas de referencia en quince idiomas, además de benchmarks comparativos repetibles frente a otros frameworks de Node.js. Forge también está pensado para asistentes de IA, con archivos de texto plano, datasets y un servidor MCP de solo lectura.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

LineLoom v1.2.0 Adds Activity Bar Control Center for Local Ollama Code Completion in VS Code

LineLoom v1.2.0, a free MIT-licensed open-source VS Code extension, is now stable on GitHub and through the Visual Studio Marketplace and Open VSX. The extension delivers AI inline code suggestions using locally run Ollama models, with optional approved remote HTTPS endpoints. The headline change is a native Activity Bar sidebar that consolidates completion status, model selection, language controls, privacy and endpoint information, and diagnostics shortcuts. The release also adds in-memory session statistics, clearer error guidance, and hardened model-selection validation. Native completions cover nine language IDs, with Go, Java, C, and C++ still labeled experimental and C# and Dart not yet implemented.

0 comments Read more at DEV Community
← NewerPage 284 of 3055Older →