SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World

Programming

LatestTopDiscussed
0
ProgrammingDEV Community ·

Developer halves checkout API latency to under 900ms using AI-assisted debugging

A software developer reduced their checkout API's p99 latency from 2.1 seconds to under 900ms in a single afternoon by using Claude Code as a debugging partner. The performance issue had gone unnoticed for about a month until a support ticket flagged that checkout felt occasionally slow. By feeding real trace data to the AI tool, the developer discovered that orders with more than 12 line items were consistently the slowest, pointing to a non-linear scaling problem. The root cause turned out to be an O(n²) bug in the discount-application logic, where every line item triggered a full rescan of all items to check for bundle discounts. An earlier attempt to fix the issue by adding a database index had failed to improve p99 latency, underscoring that tail-latency problems often require targeted fixes rather than average-case optimisations.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Builds Browser-Based Image Compressor That Never Uploads Your Files

A web developer created ImageSlim, a client-side image compression tool that processes JPG, PNG, and WebP files entirely within the browser, meaning images never leave the user's device. The tool was built in response to common frustrations with existing solutions, including privacy risks from server-based online tools, workflow disruptions from desktop software, and configuration overhead from CLI scripts. ImageSlim uses native browser technologies — the FileReader API, HTML5 Canvas, and the toBlob() method — to resize and compress images without any external dependencies or backend infrastructure. The developer addressed technical challenges such as memory limits on mobile devices for large files and correct handling of alpha transparency when converting PNGs to WebP. The project highlights the growing capability of modern browsers to perform tasks that once required server-side processing.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Prompt Injection Attacks on AI Models: How They Work and How to Defend Against Them

Prompt injection is a security vulnerability where untrusted text overrides a developer's instructions to a large language model, causing it to behave in unintended ways. The attack takes two forms: direct injection, where a user types malicious instructions, and indirect injection, where harmful commands are hidden inside documents, web pages, or emails that an AI agent reads. Indirect injection poses the greater risk for AI agents and retrieval-augmented generation systems, as attackers can plant payloads in data sources without ever accessing the application directly. The vulnerability stems from how LLMs process input — instructions and data arrive as the same undifferentiated stream of tokens, giving the model no reliable way to distinguish between them. OWASP lists prompt injection in its 2025 Top 10 risks for LLM applications, and experts note there is no single fix, only layered defenses that limit potential damage.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Dev builds rule-based emoji translator using Unicode data, no AI involved

A developer has launched Emoji Translator, a bidirectional tool that converts emoji sequences into plain-English descriptions and transforms English text into emoji variants. Rather than relying on an AI model, the tool uses versioned Unicode and CLDR reference data alongside reviewed editorial combinations and explicit rules. The decision to avoid AI was deliberate, aimed at preventing overconfident outputs that cannot be justified by the underlying data. The tool offers three emoji-to-text variants — minimal, balanced, and emoji-heavy — so users can select the appropriate tone instead of receiving a single opaque result. Available at emojitranslator.org, the project has no database or account system, and its source code and contribution guide are publicly accessible.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

AmigaDOS Developer Dr. Tim King Dies

Dr. Tim King, a key developer behind AmigaDOS, has passed away. King played a significant role in creating the operating system that powered the Amiga line of personal computers. The news was reported by amiga-news.de, a publication dedicated to the Amiga community. AmigaDOS was a foundational component of the Amiga platform, which gained a dedicated following during the late 1980s and 1990s. His contributions left a lasting impact on the history of personal computing.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

Aiden AI Board Has No Own Backend, But Key Privacy Policies Remain Unpublished

Aiden is an AI dev board that operates on a bring-your-own-provider model, routing screenshots, audio, and text to user-configured endpoints rather than any Aiden-hosted backend. The device captures a connected screen via HDMI, sends data to a multimodal model of the user's choice, and controls the target device over USB HID, with voice handled similarly through user-selected STT, LLM, and TTS endpoints. While the firmware is open source, the team acknowledges that several critical privacy details — including data retention periods, local memory deletion, encryption, and default logging — have not yet been published. Aiden also supports persistent local context, meaning data can still be stored depending on how a deployment is configured. The company cautions users to review their chosen third-party providers' data practices before using the device in sensitive workflows, as those providers form part of the overall privacy surface.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Builds Autonomous Turkish Content Site Powered Entirely by AI Agent

Turkish developer Güray has built kafa1milyon.com, a content portal that operates without human intervention using an LLM-based autonomous publishing system. The platform scans global news and scientific journals via RSS, selects topics, writes 700–1,100-word Turkish articles, and publishes them with SEO metadata and audio narrations. A virtual presenter character named Elif generates vertical promotional videos that are automatically shared to X and Instagram within daily posting limits. The system stores its logic in rule files that act as long-term memory for the AI agent, with all state managed through plain JSON files and static HTML. This article is the first in a series documenting the site's architecture and key design decisions.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

German Group Files Criminal Complaint Against Meta Over AI-Enabled Smart Glasses

A German advocacy group has filed a criminal complaint targeting Meta over its AI-powered smart glasses, according to a Reuters report from August 12, 2026. The complaint centers on legal concerns related to the device's capabilities, likely involving privacy or data collection issues. The action reflects growing scrutiny in Europe over wearable AI technologies that can capture or process information about individuals in public spaces. Germany has been among the stricter European nations in enforcing data protection standards, making it a notable jurisdiction for such a challenge. The complaint adds to broader regulatory pressure Meta faces across the EU regarding its artificial intelligence products.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

FMCSA open data lets anyone build a free weekly US trucking lead feed via public API

The Federal Motor Carrier Safety Administration publishes a publicly accessible Company Census File on the US DOT's open-data portal, data.transportation.gov, updated regularly with details on newly registered carriers. Every new trucking company that obtains operating authority must legally purchase primary liability insurance, cargo coverage, and other services, making fresh registrants high-value sales leads. The census data includes carrier name, address, phone, email, fleet size, and safety rating, queryable for free using standard HTTPS requests through the Socrata open-data API without scraping or paid subscriptions. A developer guide outlines how to filter the dataset by registration date and US state to extract only carriers with verified contact details, then schedule the query as a weekly automated feed. The approach eliminates the need for third-party lead list services, which can cost hundreds of dollars per month for comparable data.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

Chrome renders small JPEGs differently due to internal scaling behavior

A technical investigation explores why tiny JPEG images appear visually different when displayed in Google Chrome compared to other browsers. The difference stems from how Chrome handles image scaling internally at low resolutions. Chrome's rendering pipeline applies certain processing steps that alter the appearance of small JPEGs in ways that may not be immediately obvious to developers. Understanding this behavior is relevant for web developers who need pixel-accurate image rendering across different browsers.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

How to Prevent AI Agents from Losing or Duplicating Work Under Queue Pressure

Long-running AI agents frequently fail not due to model errors but because of poorly managed transitions between work states such as queued, leased, running, and completed. A relational database schema with explicit job statuses and a unique idempotency key can make these states visible and trackable. Engineers are advised to define concurrency limits per tenant and worker pool, and to persist the queue before acknowledging jobs to avoid data loss on process crashes. Workers should claim jobs atomically with short leases and verify lease ownership before each externally visible step, since expired leases can cause two workers to execute the same job simultaneously. For operations with external side effects like emails or deployments, recording intent before execution and reconciling unknown outcomes via provider APIs is essential to avoid duplicate actions.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

AI Agents Signing Up for Users Raises Legal and Product Design Questions

A debate is emerging in tech circles over whether AI assistants should be permitted to create accounts and start trials on behalf of users. When an agent completes a signup using a real person's details and email, it raises questions about whether standard terms-of-service acceptance remains valid if no human read the agreement. Proponents argue the underlying intent is genuine, since a real user initiated the request, while critics note that the person may be unaware of which product was selected or what terms were agreed to. A more clear-cut concern arises when the same agent is instructed to open dozens of trial accounts, which standard identity checks may fail to detect since each signup is technically user-initiated. No industry consensus has formed yet on how platforms should handle agent-driven signups, leaving product teams to decide individually whether to allow, restrict, or monitor such activity.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

Qwen 3 8-27B AI Model Released as Open-Weight Within Two Days

Alibaba's Qwen team released the Qwen3 8-27B language model as an open-weight model, making it publicly available on HuggingFace. The release came just two days after the model was initially announced or launched. Open-weight models allow researchers and developers to download and run the model locally without restrictions. The release was noted by the Hacker News community, though discussion remained limited at the time of posting.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

Developer Builds Self-Healing Web Scraper That Adapts to Website UI Changes

A developer has released an open-source self-healing web scraper designed to withstand website layout and CSS changes that typically break data pipelines. The tool uses a self-learning memory system to adapt automatically when target sites update their user interfaces. It has been tested across multiple e-commerce platforms including Amazon, Flipkart, and Robu.in for price comparison tasks. The project is publicly available on GitHub, where the developer is seeking community feedback and contributions.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How to pull free, real-time building permit data from US city open-data APIs

Building permits are public records that reveal who is about to spend money on construction, making them valuable early leads for contractors, suppliers, and service providers. Most large US cities publish permit data daily at no cost through official open-data portals, many of which run on the Socrata platform and expose datasets via a straightforward HTTPS-based API requiring no key for moderate use. Cities including Chicago, New York, Los Angeles, Austin, and San Francisco each have verified dataset IDs that can be queried using simple curl commands to retrieve structured JSON containing permit numbers, work descriptions, project valuations, and contractor contacts. The main technical challenge in building a multi-city feed is that each city uses different column names for the same fields, such as issue date and project value, requiring a custom normalization layer per city. Developers must also monitor for portal migrations, as some cities have silently discontinued old datasets or moved off Socrata entirely, breaking existing integrations.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Per-Agent AI Testing Misses the Real Failures Hidden in Handoffs

A developer running a three-agent AI system — comprising a planner, researcher, and critic — found that despite each agent scoring around 0.9 individually, the overall team produced wrong answers roughly one-third of the time. The root cause was information loss during handoffs between agents, such as a constraint set by the planner being silently ignored by the researcher in the next turn. The author notes that grading agents in isolation is structurally blind to these inter-agent gaps, since failures live in the seams rather than in any single turn. To fix this, the developer shifted focus to evaluating handoffs directly, asking whether constraints were preserved, whether each agent stayed within its designated role, and whether the final output remained consistent with earlier turns. The key insight is that a multi-agent pipeline is a chain, not a collection of independent functions, and its reliability depends on how faithfully each transition preserves what matters.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Builds Free In-Browser Video Dubbing Tool After AI Services Cost Too Much

A developer wanted to learn game development from a Spanish-language Godot course but found subtitles impractical for screencasts, where watching the screen and reading simultaneously is nearly impossible. Pricing out commercial dubbing services, he found Rask would cost around $540 for a six-hour course, while HeyGen offered a more reasonable $49 option, but both platforms lacked a simple one-time, no-subscription path for personal use. He built an open tool called Dub Any Video that runs locally in the browser, using a six-stage pipeline: audio extraction, transcription via Whisper, translation, text-to-speech via Piper, and audio-to-video synchronisation. The hardest engineering challenge turned out to be timing — fitting translated speech, which often runs longer than the original, into fixed video windows without distorting pitch or breaking the viewer's sync with on-screen actions. The solution combined speech speed adjustment capped at 1.3x, localised video slowdown, and allowing audio to spill into natural gaps rather than truncating it.

0 comments Read more at DEV Community
← NewerPage 188 of 1338Older →