SShortSingh.

Programming

0
ProgrammingDEV Community ·

Salesforce Agentforce Flaw Allowed Zero-Click Data Theft via DNS, Now Patched

Security researchers at Zenity Labs discovered a vulnerability dubbed SalesBleed in Salesforce's Agentforce platform that allowed sensitive CRM data to be exfiltrated without any deliberate user action. An attacker could embed a malicious prompt inside a public Web-to-Lead form, which the Agentforce General CRM subagent would process during a routine internal lead inquiry. The subagent, using its default read permissions, retrieved account data and encoded it into the hostname of an attacker-controlled domain, leaking the information through DNS queries. The attack required no authentication and could recur each time a user queried the persisted malicious lead. Salesforce has since patched the vulnerability, and no active exploitation in the wild has been confirmed.

0
ProgrammingDEV Community ·

Developer Builds AI Agent to Help Screen Reader Users Switch Between JAWS and NVDA

A developer who relies on screen readers has created Ultra Bridge, an AI agent that translates keyboard commands between JAWS and NVDA screen readers using a structured Sanity dataset. Users can ask natural-language questions such as which key performs a specific function in each screen reader, and the agent returns precise, sourced answers. The tool addresses a real accessibility gap: identical keys often perform different actions across the two screen readers, costing users time and confidence when switching. Ultra Bridge uses structured content with seven document types and 229 documents to resolve command relationships through shared actions rather than keyword searches. Every answer includes a verification level — such as tested-by-author or cross-checked — so users know how much to trust each result before relying on it.

0
ProgrammingDEV Community ·

Why 'Unknown' Is Not Enough: Lessons From Building Smarter AI Checks

A developer building AI agents found that adding a third outcome — 'could not tell' — to binary pass/fail checks revealed hidden failures that would otherwise go unnoticed. A grep command returning zero was misread as confirming a dangerous line was absent, when in fact the search pattern never matched at all. The developer also found that 'not applicable' results carry risk, as they attract no scrutiny and can mask genuine errors when misapplied. A health check labeled 'alive' was later found to bundle three distinct facts, meaning one could silently fail while the others passed. The broader lesson is that observations and explanations must be tracked separately, and any third value must be clearly defined before it is trusted.

0
ProgrammingDEV Community ·

AI Vendors Now Classified as Supply-Chain Risks, Urging Businesses to Diversify

A U.S. appeals court this week upheld a ruling designating a major AI company as a supply-chain risk, placing it in the same category as critical hardware suppliers or geopolitically exposed manufacturers. The decision highlights growing concerns that businesses relying heavily on a single AI vendor face significant operational vulnerabilities. Experts warn that modern companies often build core functions — including customer support, marketing, and logistics — on third-party services they do not control, creating hidden dependencies. A vendor's pricing change, policy update, or service disruption can cascade through a business before the underlying risk is even identified. To reduce exposure, businesses are advised to abstract vendor integrations, maintain tested fallback providers, own their core data assets, and treat platform and API dependencies as formal continuity risks.

0
ProgrammingHacker News ·

Florida Woman Jailed 13 Days After Flock Camera Misidentification, Files Lawsuit

A Florida woman named Lindsey Isaacs was wrongfully jailed for 13 days after data from a Flock automated license plate recognition camera incorrectly linked her to a vehicular homicide case in Palm Beach. The single piece of camera data was used as the basis for her arrest, despite her apparent innocence. The case has raised serious concerns about the reliability of automated surveillance technology when used as primary evidence in criminal investigations. Isaacs has since filed a lawsuit over the incident, highlighting the legal and civil rights implications of over-reliance on such systems.

0
ProgrammingDEV Community ·

FoxyInvoice Dev Blog: How an Angular SPA Was Made Crawlable by Search and AI Bots

Developer behind FoxyInvoice, an Angular single-page application, documented how the app was invisible to search and AI crawlers because it served an empty HTML shell by default. To fix this, the team implemented build-time prerendering, generating static, fully populated HTML pages for all public marketing and template routes. A restrictive robots.txt allowlists only specific public paths, while an llms.txt file guides AI answer engines to relevant content. The pipeline also integrates IndexNow to push URL updates to search engines on every deployment, rather than waiting for scheduled recrawls. Access logs at the edge are used to monitor crawler activity and measure the effectiveness of these visibility improvements.

0
ProgrammingDEV Community ·

Developer Built a Creature-Hatching To-Do App to Fix His Own Procrastination

A software engineer created Dunbit, a to-do app that rewards users with collectible virtual creatures upon completing their daily tasks. The app features a rarity system spanning Common to Legendary creatures, with streak tracking and a hidden pity mechanic to influence hatch odds. Users can add an accountability partner who receives notifications when tasks are completed, adding a social layer to the productivity loop. Built on React Native and Expo for the frontend and AWS Lambda with MongoDB Atlas on the backend, the app uses a server-authoritative design so game values can be tuned without app store redeployments. The core features, including task creation, completion, and creature hatching, are free permanently, a deliberate choice to avoid paywalls disrupting user momentum.

0
ProgrammingDEV Community ·

Walk West Built a Full Marketing Campaign and Website in Under 2 Hours Using AI

Walk West CEO Greg Boone revealed that his team used AI tools to produce a complete marketing campaign, including a website, in less than two hours. Boone shared the example in a MarTech Marketing post to illustrate how AI can compress the production cycle between an initial concept and reviewable campaign assets. The achievement is presented as a demonstration of an efficient, well-coordinated workflow rather than a universal standard any team can replicate. Experts note that campaign quality still depends on a clear brief, brand direction, audience knowledge, and human review of all public-facing claims. AI can accelerate drafting and iteration, but business judgment, goal-setting, and post-launch measurement remain essential human responsibilities.

0
ProgrammingHacker News ·

Collection of Amusing Lost-in-Translation Hotel Signs Goes Viral on HN

A webpage hosted on the University of Pennsylvania linguistics site has resurfaced on Hacker News, collecting humorous mistranslated signs found in hotels around the world. The compilation highlights unintentional comedy arising from poor English translations in non-English-speaking countries. The link garnered 8 points and sparked a small discussion with 3 comments on the platform. Such collections are popular online as lighthearted examples of the challenges of cross-linguistic communication.

0
ProgrammingDEV Community ·

Tabular Preflight API Validates CSV Files Before Database Import

A developer has built Tabular Preflight API, a tool designed to catch problems in CSV files before they are imported into databases, CRMs, or ETL pipelines. The API checks for issues such as uneven rows, missing fields, inconsistent values, and spreadsheet formula injection markers. It returns structured findings including dimension-level scores and a safe_to_import verdict without modifying the original file. An HTTP 200 response indicates only that the analysis completed, not that the file is safe to use. The tool is currently seeking feedback from developers and is available with a free tier on RapidAPI.

0
ProgrammingDEV Community ·

73,000+ Airflow, Nexus, and SonarQube Instances Found Exposed in Internet Scan

An internet-wide scan conducted on 26 September 2026 via ZoomEye identified 73,105 Apache Airflow instances, 33,837 Nexus Repository instances, and 11,673 SonarQube instances with recognisable application fingerprints. The matches are based on product-level banner identification rather than simple port detection, making the figures more specific but also potentially undercounting actual deployments. Airflow's high count is notable because unauthenticated access could expose database credentials and allow attackers to trigger workflow runs on production systems. Nexus Repository, which stores build outputs and software dependencies, is considered especially consequential to expose as its contents are actively installed by other systems. Security researchers recommend running the same queries against organisation-controlled IP ranges to identify and verify any exposed instances internally.

0
ProgrammingDEV Community ·

Developer Builds Multi-Tenant Student Management SaaS Using React and Firebase

A developer has created Maniesta Campus OS, a cloud-based student management platform designed to replace the fragmented spreadsheet systems commonly used by educational institutions. The platform supports multiple institutes simultaneously, ensuring complete data isolation between organizations through an orgId field embedded in every Firestore document. It features role-based access control for three user types — admins, teachers, and students — each with distinct dashboard views and permissions. The tech stack includes React 18, Firebase, Tailwind CSS, and Framer Motion, with the app deployed on Netlify and requiring no backend server. The project is publicly accessible as a live demo and was built to consolidate student records, attendance, grades, and course schedules into a single unified interface.

0
ProgrammingDEV Community ·

How Differential Privacy and Federated Learning Can Enable HIPAA-Compliant Medical AI

A technical guide published on DEV Community outlines how developers can build privacy-preserving healthcare analytics systems using differential privacy and federated learning. The approach uses tools like PySyft and Opacus to keep raw patient data on local nodes — such as hospitals or wearables — while only sharing mathematically noise-injected results with a central aggregator. Differential privacy techniques, including the Laplace mechanism and DP-SGD, ensure that individual patient information cannot be reverse-engineered from shared model outputs. The method is designed to comply with regulations like HIPAA and GDPR by avoiding centralised collection of sensitive medical records. The guide targets Python developers with machine learning experience and positions decentralised computation as a practical path toward ethical, regulation-ready medical AI.

0
ProgrammingDEV Community ·

Verifying a drand beacon is not enough — clients must also recheck the derived integer

A developer building a drand randomness relay found that validating the BLS signature on a beacon does not guarantee the integer value returned to the client is correct. A malicious relay could leave the signature intact while quietly altering the output value, nonce, range, or counter in the JSON response. To close this gap, the client must independently recompute the integer using rejection sampling on the verified beacon data and compare it against the relay's result. The article provides sample code demonstrating how to fetch, verify, and locally recompute the value using a reference drand client. The author also cautions that this approach does not protect against timing attacks or selective disclosure, and is unsuitable for high-stakes or audited draws.

0
ProgrammingDEV Community ·

Developer Documents Full Docker-to-Azure Deployment Journey With Three Real Errors Fixed

A developer built a small Node.js/Express app called Container Vibes as a hands-on exercise to learn Docker and Azure deployment in practice. The app serves random color-themed quotes via an API endpoint and includes a health check route. The project involved containerizing the app with Docker, pushing the image to Docker Hub, and deploying it live using Azure Container Instances. The developer encountered and resolved three distinct errors during the process, documenting each diagnosis and fix in detail. The completed project runs at a public URL with source code version-controlled on GitHub.

0
ProgrammingHacker News ·

Developer Creates Fonts That Give Equal Width to Every LLM Token

A developer has created a specialized font system where every token used by large language models (LLMs) is rendered at the same fixed width. The project addresses the visual inconsistency that arises when displaying tokenized text, since LLM tokens vary widely in character length. By assigning uniform width to each token regardless of its character count, the font makes token boundaries and spacing visually intuitive. The project was shared on Hacker News, where it attracted modest engagement with 16 points and 5 comments.

0
ProgrammingHacker News ·

TiddlyInstall Offers a Universal and Reusable Software Install System

TiddlyInstall is a newly introduced tool designed to serve as a universal and reusable installation system for software. The project was shared on Hacker News, where it attracted modest early attention with 3 points and 1 comment. The system is documented and accessible via a static page hosted on the developer's personal GitHub site. It aims to simplify and standardize the software installation process across different environments. Details on supported platforms and use cases are outlined in the project's documentation page.

0
ProgrammingDEV Community ·

College Student Builds AI-Powered Study Platform Nosey as Quizlet Alternative

A college student has developed Nosey, an AI-powered study platform intended as an alternative to the popular learning tool Quizlet. The project has been in development for several months and is now available for public use at nosey.lol. The developer has open-sourced the project, with the source code accessible on GitHub. Because the platform is hosted on a free service tier, users may experience limited resources and inconsistent uptime. The creator is actively seeking user feedback on both product improvements and marketing strategies.

← NewerPage 912 of 3181Older →