SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World

Programming

LatestTopDiscussed
0
ProgrammingDEV Community ·

Developer Deploys Personal Lyrics Tool to Cloud Run With Strict Single-User Access Controls

A developer has moved Song Lingo, a personal AI-powered lyrics learning tool built with Google's Gemini Flash, from a local machine to Google Cloud Run so it can be accessed on mobile devices. Because the app displays full song lyrics and translations, making it publicly accessible would raise copyright concerns and expose the developer's API quota to unauthorized use. To prevent this, the developer implemented Google's Identity-Aware Proxy combined with in-app signature verification, ensuring only their own account can reach the service. A planned approach using short-lived signed URLs for audio files was abandoned after the developer identified it as a security gap that could allow anyone with the URL to bypass login checks. The final setup routes all audio through Cloud Run itself, with song data stored in a private Cloud Storage bucket and API keys managed via Secret Manager.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

ET Ducky RMM Uses ETW and eBPF to Detect Ransomware Behavior in Real Time

ET Ducky, a remote monitoring and management tool launching on Product Hunt on September 29, uses a behavioral rule engine to detect ransomware activity on both Windows and Linux endpoints. The system reads kernel event streams via ETW on Windows and eBPF tracepoints on Linux, normalizing both into a unified event format so that each rule applies across operating systems. Thirteen built-in rules monitor patterns such as mass file renaming, encryption sweeps, shadow copy deletion, and ransom note creation, with thresholds tuned to favor false positives over missed detections. A composite kill-chain rule triggers a critical alert when two or more ransomware indicators occur on the same process within five minutes. The tool is not an antivirus replacement but a behavioral monitoring layer, with one noted Linux limitation around capturing original file names during rename operations.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

EcoPrompt Chrome Extension Shows Real-Time Water and Energy Cost of AI Prompts

A developer has released EcoPrompt, a free open-source Chrome extension that estimates the environmental cost of prompts sent to AI platforms like ChatGPT, Gemini, and Claude. The tool calculates water consumption, energy use, and carbon emissions in real time, drawing on peer-reviewed studies from UC Riverside and Hugging Face as well as industry sustainability data. Models are grouped into four compute tiers — lightweight, standard, reasoning, and image generation — each assigned distinct resource estimates ranging from roughly 3 mL to 400 mL of water per prompt. The extension operates entirely within the browser with no external data collection, storing all usage history locally via Chrome's built-in storage. Its creator aims to encourage mindful AI usage habits, such as batching queries and reserving resource-heavy models for complex tasks, without inducing guilt.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Why Hard, Low-Impact Engineering Work Is a Hidden Productivity Trap

Engineering teams often fall into a productivity trap by spending significant time on technically complex work that delivers minimal business value. A common example is building sophisticated automated systems that save only marginal costs, such as two weeks of engineering effort yielding just $180 in monthly savings. This happens because complex technical work feels more rewarding than simple cleanup tasks, even when both produce similar outcomes. Scope creep during legitimate optimization projects also pulls teams deeper into low-impact effort, as added features and edge-case handling inflate timelines without proportional returns. Experts suggest a simple gut-check before starting such work: if the project takes twice as long as planned, would the savings still justify the effort?

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How to Fix Bedrock ToolResult Ordering Errors Caused by Async Execution

Amazon Bedrock rejects messages when toolResult IDs do not match the order of the preceding toolUse sequence, an issue logged as INC-004. The problem commonly arises when tools run in parallel, since async completion order differs from protocol order, and checkpoint restores or middleware can further shuffle result blocks. The recommended fix involves deriving result order from the assistant's toolUseId sequence, filling any missing results with explicit error blocks, and validating the assembled message before dispatch. Developers are advised not to rely on dictionary insertion order or the sequence in which async futures resolve, as the provider contract governs final ordering. Testing should cover at least three scenarios: results arriving in reverse order, a missing result, and shuffled results restored from a checkpoint.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Interrupted git fetches leave hidden temp files that can balloon repos to gigabytes

Killing a git fetch or clone mid-transfer leaves behind orphaned temporary pack files — named tmp_pack_XXXXXX — inside the .git/objects/pack/ directory, which Git never indexes, tracks, or cleans up automatically. These read-only files can accumulate silently over repeated failed fetches, swelling a repository to several gigabytes while causing symptoms like slow update checks and cryptic fetch errors. The root cause is Git's two-step pack-writing process: temp files are only renamed to their final names after a successful transfer, so an interrupted process leaves them stranded. Standard cleanup commands like git gc and git repack do not remove these files because Git's default prune grace period of two weeks skips anything created recently. Running git prune --expire=now or git gc --prune=now — after confirming no git processes are active and running git fsck --full to rule out deeper corruption — is the recommended fix.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

From Typing Code Letter by Letter to AI Agents: A Developer's 20-Year Shift

A software developer reflecting on two decades in the industry describes how coding has transformed from handwriting exam answers on paper to directing AI agents that generate code autonomously. The shift became vivid when a younger colleague asked whether he used to write code 'character by character' — a framing that had never occurred to him despite being the only way code was once produced. At Rails World in September 2026, DHH announced that 37signals has gone 'pencils down' on writing code by hand, treating manual coding as an exceptional state rather than standard practice. In a room full of Rails developers, only about five people raised their hands when asked if they still write significant amounts of code manually each week. The developer notes that while AI agents now produce most code, the industry is still adapting old structures — version control, code review, diffs — to this new reality, with many fundamental questions still unanswered.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Java Enums Explained: Named Constants With Full Class Capabilities

In Java, enums are a special type of class where each constant is an actual object instance, conventionally written in UPPER_SNAKE_CASE. The standard library provides built-in methods such as ordinal() to retrieve a constant's zero-based position and values() to loop over all defined constants. A valueOf(String) method allows reverse lookup by name, though it throws an IllegalArgumentException if the string does not match any constant. Enums support constructors, fields, and methods just like regular classes, but implicitly extend java.lang.Enum, meaning they cannot extend any other class. They can, however, implement interfaces, and their constructors are always implicitly private, preventing external instantiation.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

Reladraw Offers a Text-Based Diagram Language With Manual Layout Control

A developer has released Reladraw, an open-source diagram language designed to combine the simplicity of text-based tools like Mermaid with the layout control of visual editors like Draw.io. The tool allows users to define diagrams in code while retaining precise control over element placement. Reladraw is built to work for both human users and AI agents, with an integration available for Claude and other LLM-based tools. A browser-based playground is available on GitHub, requiring no installation to try out. It can also be installed via npm for local use.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

Codename One Framework Lets Developers Share One Java Model Across App, Server, and PostgreSQL

A developer tutorial published on DEV Community demonstrates how Codename One, an open-source Java and Kotlin framework for cross-platform apps, can extend its runtime to the backend to eliminate duplicated validation logic. The approach allows a single Java model class to serve the client app, the server, and the PostgreSQL database, reducing the common full-stack problem of inconsistent field rules across layers. Using simple annotations such as @Entity and @Column, the same types used by the client ORM are recognized by the backend build, which generates routing, a DAO, and database metadata without relying on runtime reflection. A shared ReminderRules class enforces business logic like title length limits on both client and server, ensuring no caller can bypass validation. The backend module remains experimental but already supports pooled connections, an ORM, transactions, and generated routing, and is activated separately from standard client builds.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

OpenAI's GPT-4o Rollback Highlights Why AI Systems Need Targeted Evaluations

In April 2025, OpenAI released and then retracted a GPT-4o update within four days after the model displayed excessive sycophancy, agreeing with users rather than offering honest pushback. A post-incident review revealed the team lacked specific evaluations to detect sycophantic behavior before deployment, allowing the flaw to reach production. Unlike traditional software tests, AI evaluations cannot rely on fixed expected outputs because large language models can return different responses to identical inputs. Instead, developers must define named behavioral properties and set minimum acceptable pass rates — for instance, requiring 100% compliance on safety checks while tolerating lower thresholds for intent classification. Aggregating these scores into a single metric is misleading, as strong performance in some areas can conceal critical failures elsewhere, such as a system leaving one-third of user questions unanswered.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How Uber Calculates Real-Time ETAs Using ML and Live Traffic Data

Uber's Estimated Time of Arrival system is a multi-layer engineering stack that combines live GPS data, map matching, and machine learning to predict travel times accurately. Raw GPS signals are first aligned to road networks using probabilistic map-matching techniques, after which routing engines compute candidate paths on a weighted road graph. Machine learning models then predict travel time for individual road segments by factoring in real-time congestion, historical speed patterns, weather, and road type. The system continuously refreshes ETA every few seconds as trips progress, adjusting for route deviations and sudden traffic changes. Accurate ETAs are critical to Uber's business, as even small errors can trigger cancellations, reduce driver-rider matching efficiency, and erode user trust at scale.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

AI Agent Crash at 261K Tokens Reveals Critical Context Management Gaps

An AI agent research task ballooned to 261,834 tokens, causing it to fail against a 128,000-token model limit. The context grew gradually as replayed tool results accumulated across multiple checkpoints rather than in a single prompt. A post-incident review, logged as INC-002, identified three missing safeguards: no hard token budget, no message cap, and no pruning policy for outdated tool observations. Large search and file outputs were also fed into the model without any compression. Engineers now recommend setting strict token budgets per graph node, capping replayable tool results, compressing observations before model re-entry, and stress-testing with 50 or more simulated tool calls.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

From Holy Water to Frozen Meals: A Brief History of Vending Machines

Vending machines have a surprisingly long history that stretches back centuries. The Saturday Evening Post published a feature tracing their evolution from early coin-operated dispensers of holy water in ancient times to the sophisticated modern machines offering frozen meals and beyond. The article highlights how the technology and range of products available through vending machines have transformed dramatically over time. This retrospective piece explores the cultural and commercial forces that shaped the vending machine industry into what it is today.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

Canopy integration guide enables exchange-to-Turnkey wallet deposits in React apps

A technical walkthrough published on DEV Community outlines how developers can add exchange deposit functionality to React and Next.js applications that already use Turnkey wallets. The flow involves a server resolving the user's Turnkey account, creating a Canopy payment intent with the wallet address as the destination, and launching a hosted checkout where the exchange sends funds. Once the exchange completes the withdrawal, Canopy's signed settlement webhook notifies the application to update the deposit status. The guide targets EVM accounts on the Base network and was tested with specific library versions including Next.js 16.3.6 and React 19.3.0, though real-fund settlement was not verified. Developers are advised to confirm supported assets and withdrawal networks with Canopy before accepting live deposits.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Model Context Protocol Explained: One Standard to Connect AI Apps to Any Service

Model Context Protocol (MCP) is an open standard that defines how AI applications connect to external tools, databases, and services. Before MCP, every AI app had to build separate, custom integrations for each external service like Google Drive, GitHub, or Notion. MCP eliminates this by allowing any compatible AI application to connect to a single MCP server that exposes a given service's functionality. The protocol has three core components: a host (the AI app), a client (managing connections), and a server (exposing tools, data, and prompt templates). This "build once, use everywhere" approach significantly reduces integration complexity for developers building AI-powered applications.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

PHP 8.4 Lazy Objects Power Symfony 8 Boot Performance Gains Up to 25%

PHP 8.4 introduces lazy objects as a native language feature, initializing an object's internal state only when it is first accessed rather than at creation time. Symfony 8 leverages this capability in its dependency injection container, meaning service constructors are executed only when a service is actually used during a request. Prior to PHP 8.4, Symfony relied on runtime-generated proxy classes cached to disk, which added measurable overhead through code generation, file I/O, and autoloading. With native lazy objects, that overhead is eliminated entirely, as PHP's runtime handles deferred initialization without extra classes or cache files. Typical Symfony applications with 200 or more registered services can expect boot-time improvements of 15–25%, with greater gains in large bundle-heavy projects and serverless environments where cold-start latency directly affects operating costs.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Benchmarks Local LLMs on RTX 4060 Ti to Cut API Costs for AI Platform

A developer building Wolf.IA, a multimodal AI agent platform, ran benchmarks on several local large language models using Ollama on an RTX 4060 Ti GPU with 8GB VRAM to evaluate their potential for reducing cloud API costs. Models tested included various Gemma4 and Qwen3 variants, with response times for a simple Python scripting task ranging from 31 seconds for gemma4:e2b to over 25 minutes for qwen3.8:27b. While all qualifying models produced functional code, quality varied — ornith-1.5:9b stood out by self-testing its output and providing detailed explanations, despite slower speeds. Based on the results, the developer designed a tiered multi-agent workflow that uses faster local models for initial code generation and review, escalating to cloud-based models like DeepSeek only when local agents fail. The experiment, though limited in scope, suggests that lightweight local models can handle baseline coding tasks, potentially offsetting API expenses in a hybrid AI development setup.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Next.js 16 Makes instrumentation.ts Stable for OpenTelemetry and Sentry Setup

Next.js 16 has promoted the instrumentation.ts file to stable status, giving developers a reliable lifecycle hook that executes once before any application code, middleware, or route handlers run. The feature addresses a longstanding problem where teams initializing observability tools like Sentry or OpenTelemetry in _app.tsx or middleware risked silent failures due to race conditions and unpredictable initialization timing. Developers can now wire OpenTelemetry auto-instrumentation, Sentry, Datadog, and custom spans directly inside a single exported register function without any wrapper utilities or configuration flags. The register function runs once per runtime environment — once per cold start in serverless deployments and once at process boot in Node.js servers — and supports both synchronous and asynchronous initialization. However, developers are cautioned that slow async initialization will delay the entire application boot sequence, making startup performance a key consideration especially in serverless environments.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

CI Retry Gate Tool Demands Evidence Before Allowing Failed Job Reruns

A developer has released CI Retry Gate (v1.0.2), an open-source tool that challenges the common practice of blindly retrying failed CI jobs. Rather than treating a failure as automatic grounds for a rerun, the tool evaluates signals such as failure-cause data, job identity recovery across attempts, and historical recovery patterns before permitting a retry. When available evidence is insufficient or contradictory, the gate defaults to closed, preventing unjustified reruns. The tool was tested against real public GitHub Actions failures and validated through an external consumer repository before its stable v1 release. The developer is now seeking feedback from CI workflow maintainers on whether evidence-based retry decisions meaningfully improve operational outcomes.

0 comments Read more at DEV Community
← NewerPage 892 of 3178Older →