SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World

Programming

LatestTopDiscussed
0
ProgrammingDEV Community ·

Developer Builds Open-Source Tool to Route Campus Firewall Blocks for Low-Lag Gaming

A developer has released an open-source Windows tool called AetherRoute that allows students to bypass campus firewall restrictions on gaming traffic without using a traditional VPN. The solution works by simultaneously connecting to campus Wi-Fi and a smartphone USB tether, then manipulating Windows network interface metrics to route different traffic types through different connections. Authentication, DNS queries, and in-game voice chat are directed through the mobile data connection, while high-bandwidth match traffic is routed directly through the campus Wi-Fi for low latency. A background UDP socket sniffer dynamically identifies active game server IP addresses and injects destination-specific routes in real time via PowerShell, eliminating the need for manually maintained server lists. The approach reportedly preserves the low ping of campus infrastructure while avoiding the jitter and anti-cheat flags associated with conventional VPN tunnels.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Security blogger walks through HTB Forest machine ahead of PJPT certification exam

A cybersecurity student preparing for the TCM Security PJPT exam documented their experience tackling Hack The Box's Forest machine, a Windows Server 2016 Active Directory target. The machine, rated 'Easy' on the platform, is widely regarded as a practical environment for learning Active Directory enumeration and attack techniques. The writeup covers initial reconnaissance using Nmap, which revealed key services including Kerberos, LDAP, SMB, and DNS pointing to the domain htb.local. Despite getting stuck at several points, the author completed the machine and gained hands-on experience with AD-based attack methodologies. They advised that basic Active Directory knowledge is recommended before attempting the machine, as beginners may find it difficult to follow without that foundation.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer releases open-source SDK to make React Native apps resilient on poor connections

A developer has released NetResilience, an open-source SDK for React Native designed to improve HTTP communication reliability under poor internet conditions. Unlike standard connectivity libraries that only detect online or offline status, NetResilience classifies network quality into four levels: Excellent, Moderate, Poor, and Offline. Based on the detected quality, the SDK can queue, retry, delay, or reject HTTP requests according to configurable resilience policies. Pending requests are persisted locally using expo-sqlite, allowing queued operations to survive app restarts. To address security concerns, the SDK strips authentication tokens before local storage and fetches a fresh token via an authProvider callback at sync time, preventing issues with expired credentials.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

OpenAI Cuts GPT-5.6 Prices and Adds Fast Mode to Expand API Options

OpenAI has announced pricing and performance updates to its GPT-5.6 model lineup, reducing Luna's cost by 80%, Terra's by 20%, and introducing a Fast mode for Sol. The company frames these changes as an effort to advance what it calls a price-performance frontier, giving developers more tailored options across different workloads. CEO Sam Altman has publicly stated the goal of offering the best intelligence-to-cost ratio at multiple points along a trade-off curve, rather than relying on a single flagship model. The updates signal a shift toward more operational model selection, where cost, speed, and output quality are weighed together depending on the use case. OpenAI has not specified a universal formula for choosing between its GPT-5.6 variants, leaving teams to evaluate the options against their own production requirements.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Automates Bug Reports from Tally Forms to GitHub and Discord via Make

A developer building a privacy-focused Android app called Timety needed a way to collect user feedback without requiring account creation. To avoid manually handling form submissions, they built an automated workflow connecting the form tool Tally, GitHub, and Discord using the automation platform Make. The setup routes three types of submissions — bug reports, feature requests, and general feedback — to GitHub Issues or Discussions with appropriate labels, while also sending alerts to a Discord channel. Bug reports and feature requests use GitHub's GraphQL API to populate existing issue templates, with an extra step to format any screenshot URLs for Markdown. The workflow provides a fallback since Tally still sends email copies of submissions if the automation fails.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Hacktron AI Researchers Hack OpenAI via Forum Exploit, Earn $6,500 Bug Bounty

On September 18, Ars Technica reported that researchers from Hacktron AI successfully compromised OpenAI systems using a multi-step attack chain. The breach began by exploiting a vulnerability in OpenAI's third-party Discourse forum, then escalated through internal sign-on systems to access an employee's ChatGPT account. Attackers ultimately exfiltrated sensitive code from GitHub and proposed unauthorized code changes before disclosing the findings. OpenAI acknowledged the researchers' report and awarded a $6,500 bug bounty for the discovery. The incident highlights how sequential, low-profile access steps can evade conventional single-request security scanners that do not analyze attack chains across multiple interactions.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

ChaosCypher v0.4.3 Fixes Four Silent Queue Bugs and an Arbitrary File Read Flaw

ChaosCypher released version 0.4.3 with no new features, instead addressing 23 code-level fixes across 54 commits since v0.4.2. Four queue-recovery defects were resolved, including a counterintuitive bug where a graceful worker shutdown permanently killed recoverable tasks while a hard crash left them retrievable. Other fixes corrected a double-counting retry budget error, a heartbeat refresh that silently failed on expired keys, and a race condition that allowed duplicate chunk processing and billing. A security flaw in the MCP add_document tool was also patched, as it bypassed sandbox containment when content was supplied, potentially exposing sensitive files like credentials. Additionally, justification text stored alongside graph edges is now sanitised to remove model reasoning artifacts, ensuring only genuine evidence is shown to users.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

OpenAI Slashes GPT-5.6 Luna API Price by 80%, Terra Cut by 20%

OpenAI has significantly reduced API pricing for two of its GPT-5.6 models, Luna and Terra, making high-volume AI workloads more affordable for developers and businesses. Luna now costs $0.20 per million input tokens and $1.20 per million output tokens, reflecting an 80% price reduction, while Terra is priced at $2 and $12 per million input and output tokens respectively, a 20% cut. The changes are rolling out across platforms including AWS, and both models remain accessible via ChatGPT Work, Codex, and the OpenAI API. OpenAI also retains a Fast mode for GPT-5.6 Sol, offering up to 2.5 times faster speeds at twice the standard price, automatically applied to priority requests. The lower token costs are expected to make previously cost-prohibitive automation use cases more viable and reduce expenses for existing workflows without requiring engineering changes.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

OneTUI: One Terminal UI to Browse Postgres, Kafka, NATS, and Qdrant

A developer has released OneTUI, an open-source terminal user interface tool designed to provide a single, consistent interaction model across multiple databases and message queues. The project currently supports Postgres, Kafka, NATS, and Qdrant, allowing users to browse data sources using unified keybindings and navigation patterns. OneTUI draws inspiration from K9s, the Kubernetes CLI tool, adopting its resource-browser model and applying it to a broader range of data sources. The tool is built in Rust using the ratatui library and is currently limited to read-only mode, with write support planned once the core interaction model matures. The source code is available on GitHub at github.com/syndbg/onetui, and the developer is actively seeking feedback on the concept.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Five Governance Questions Microsoft Says Enterprises Must Ask Before Deploying AI Agents

As AI agents built with Microsoft Foundry move into everyday business use, enterprises face pressing governance challenges around ownership, access control, and accountability. Microsoft recommends teams answer five key questions before a broad rollout: who owns the agent, who can invoke it, which actions require approval, how incidents can be investigated, and how the agent can be disabled if needed. The Foundry Control Plane provides a subscription-level inventory of deployed agents, while the Microsoft Agent 365 Registry offers a broader organization-wide view for agents built across platforms like Copilot Studio. Access governance involves two distinct checks — one controlling who can call the agent and another governing what data or systems the agent itself can reach. Enterprises are also advised to pin agents to approved versions in production, since the default behavior serves the latest version, which could introduce unreviewed changes.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer details how a simple AI brand tracker breaks down at scale

A developer building an AI visibility tracker for CrawlSpider shared how a basic Python script that checks whether a brand appears in LLM responses quickly becomes inadequate at scale. While a few lines of code suffice for tracking one brand across a handful of prompts, scaling to 100 brands, 50 prompts, and 3 models generates 15,000 API requests daily. At that volume, the architecture must evolve to include job queues, worker pools, retry logic, rate-limit management, and idempotency controls. Simple string matching also proves unreliable, as brand names like 'Apple' can be ambiguous, prompting the need for entity extraction and competitor classification. The developer further noted that tracking rank changes over time — not just binary mentions — is what makes the data genuinely useful.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer builds cross-platform link-saving app with AI summaries and device sync

A developer built LinkVault Professional, a personal link-management app, after failing to find an existing tool that stored articles, videos, and repositories with context like titles, summaries, and categories. The app uses the operating system's native Share function for quick saving and optionally employs an AI model to auto-fill metadata. It now runs across Apple iOS, macOS, Android, Windows, a Chromium/Edge extension, and a progressive web app, with cross-device sync tied to user accounts. Multilingual support is built in, with categories served centrally from the server and displayed in the user's chosen language. Following user feedback, the app also supports multiple linked logins sharing a single library across devices.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Fanless AI Servers Require Near-Total Liquid Cooling Beyond 250 kW Per Rack

As AI server rack densities surpass 250 kW, traditional hybrid liquid-air cooling systems become inadequate, leaving tens of kilowatts of heat that air infrastructure struggles to manage. The emerging solution is near-total liquid heat capture, which reduces residual air cooling needs to under one percent and allows fully fanless server operation. Rising processor thermal design power is spreading heat to previously air-cooled components such as memory, networking, storage, and power modules, all of which now require dedicated liquid cooling. CoolIT has been developing modular coldplate-based cooling loops across six generations of fanless server designs to address these increasingly complex thermal demands. Engineers building next-generation AI servers must now design comprehensive liquid cooling strategies that account for every heat-generating component on the motherboard, not just CPUs and GPUs.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Go Startup Credential Check Catches Misconfigs Before First Request Hits Users

A recommended practice for B2B SaaS deployments advises verifying a service credential's accepted identity at startup rather than waiting for a live request to expose a misconfiguration. Running a pre-promotion identity check converts a wrong-key scenario into a deploy failure instead of a customer-visible error. The approach involves calling an authenticated endpoint, comparing the full JSON response against a reviewed expected document stored in deployment configuration, and recording the release identifier, check time, and outcome without logging the credential itself. A small Go program illustrates the pattern, implementing retry logic with exponential backoff for rate-limited responses and a bounded response read. The method is acknowledged as a point-in-time snapshot, meaning runtime credential revocations still require separate alerting and error handling during normal operation.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

SQL Window Functions vs. Aggregate Functions: Key Differences Explained

SQL aggregate functions like SUM() and COUNT() collapse multiple rows into a single summary row per group, causing individual row details to be lost. Window functions perform similar calculations but retain all original rows, attaching computed results to each one using the OVER() clause. Unlike aggregate functions, window functions support ranking operations such as RANK() and DENSE_RANK(), enabling per-group rankings without complex subqueries or self-joins. Adding an ORDER BY clause inside OVER() further allows running totals and cumulative sums, a capability aggregate functions cannot replicate. The core distinction is that aggregate functions summarize data while window functions annotate it, making the latter more flexible for analytical queries.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Builds Excel Dashboard to Analyse Jumia E-commerce Product Data

A developer analysed 115 Jumia product records as part of a personal data analysis case study. During the data cleaning phase using Python and Pandas, three records with invalid review values were removed, leaving a final dataset of 112 products. The cleaned data was then loaded into PostgreSQL, where SQL queries were used to explore business questions around product discounts and pricing. Finally, the structured data was imported into Microsoft Excel, where pivot tables and summary charts were built to visualise patterns in discounts, ratings, and review activity. The full project, including code and findings, has been published on GitHub.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

OwlLayer launches open-source SDK for building AI-powered UI across major frameworks

OwlLayer is a newly released open-source TypeScript SDK designed to help developers build AI-powered user interfaces. It supports multiple popular frameworks including React, Vue, Angular, Svelte, Vanilla JS, and PHP. The SDK includes features such as voice interaction, live context awareness, and Human-in-the-Loop (HITL) security controls. It is built with MCP integrations in mind, targeting WebMCP and NativeMCP use cases. The project is publicly available on GitHub under the handle borisbob91.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

A Two-Button Ring With No App Outperforms Every Smart Ring Its Owner Tested

A wearable technology researcher bought a Bluetooth ring equipped with only two physical buttons and a glide sensor, paying more for it than any other ring in his collection. Unlike smart rings that promise health monitoring, AI features, and sleep tracking, this device simply connects via BLE and turns the finger into a mouse cursor controller. The ring paired instantly with no account creation, firmware prompts, or location permissions required. When sweat caused false inputs on the glide sensor, powering the ring off and back on with a five-second button press resolved the issue immediately — a fix impossible on capacitive-touch devices. The researcher argues that physical buttons represent a fundamental design commitment to the user's real-world conditions, and identified seven usable functions from just two buttons and the glide surface alone.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

AI Can Screen Resumes, But Human Judgment Still Matters in Hiring

Artificial intelligence is increasingly being used to support recruitment tasks such as resume screening, candidate matching, assessments, interview scheduling, and communication. These tools can free up recruiters' time by handling high-volume, repetitive parts of the hiring process. However, experts argue that evaluating how a candidate thinks, approaches problems, or fits a team culture still requires human judgment. AI lacks the ability to provide context or ask nuanced follow-up questions the way a human interviewer can. The growing use of AI in HR technology has sparked broader debate about where automation should end and human decision-making should begin.

0 comments Read more at DEV Community
← NewerPage 1009 of 3192Older →