SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World

Programming

LatestTopDiscussed
0
ProgrammingDEV Community ·

Qwen-Image 2.1 Tested on 48 GB MacBook Pro via ComfyUI: Speed, Edits, and Limits

Qwen released its Image 2.1 model on September 20, 2026, offering a unified 7-billion-parameter system for image generation, editing, transparent output, and multi-reference support. A developer ran 20 local generations on a MacBook Pro with 48 GB unified memory using ComfyUI Desktop, covering baseline images, transparent assets, edits, posters, and higher-resolution outputs. Median generation times were approximately 143.5 seconds at 1024×1024 and 343.5 seconds at 1536×1536, with results often appearing convincing despite missing some placement, text, or color-consistency requirements. The model is released under the Qwen Research License Agreement, which restricts use to non-commercial purposes such as research and evaluation, requiring a separate license for any commercial deployment. The evaluation highlights that download access, local execution, and commercial permission are distinct considerations that users must address independently before building production pipelines.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer's AI-run Stephen King site published fabricated quotes for months

A developer running a Spanish-language Stephen King fan site automated nearly all content creation using an AI agent that publishes articles every four hours without human review. After six months of unattended operation, audits revealed serious accuracy failures: 34 of 90 quotes were invented, paraphrased, or misattributed, including lines falsely credited to King that actually belonged to Faulkner, Lord Chesterfield, and others. Fifteen of 16 featured book quotes in the CMS did not exist, and one Cujo quote inverted the meaning of the real sentence entirely. The developer noted the errors were especially hard to catch because fabricated content was optimised to sound more thematic and better-cadenced than real quotes, meaning standard human review would likely have passed them. The experience led to a key editorial rule: a quote is only usable if the source can be named explicitly within the text itself.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

OpenAI Formalizes Deeper Third-Party Safety Testing for Frontier AI Models

OpenAI has expanded its external AI safety testing program, giving qualified third-party organizations more structured and deeper access to evaluate its frontier models across training, evaluation, and deployment stages. The initiative covers three forms of collaboration: independent capability evaluations, methodology reviews, and domain-expert probing of specialized risk areas. For its GPT-5 generation, OpenAI coordinated assessments spanning areas such as long-horizon autonomy, deception, oversight subversion, and offensive cybersecurity. Assessors may receive secure access to early model checkpoints, reasoning traces, and results from testing with fewer safety mitigations — access distinct from what standard API customers receive. OpenAI frames external testing as one layer within a broader safety ecosystem, with findings disclosed publicly through system cards and collaborator publications.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

Opinion: Has JavaScript Lost Its Way Amid Growing Complexity?

A blog post by developer Maroun Baydoun explores what he calls a 'midlife crisis' facing JavaScript as a language and ecosystem. The piece reflects on how JavaScript has evolved significantly from its early simplicity to a complex landscape of frameworks, tooling, and competing standards. Baydoun suggests this rapid evolution may be causing confusion and fatigue among developers. The post sparked discussion on Hacker News, drawing comments on the state and direction of modern JavaScript development.

0 comments Read more at Hacker News
0
ProgrammingHacker News ·

Developer Demos Native Apps Built Using TypeScript and CSS

A developer has shared a collection of example projects on GitHub showcasing native applications written in TypeScript and CSS. The repository, hosted under the account 'geastack', aims to demonstrate that native app development is possible using web-familiar technologies. The project was submitted to Hacker News, where it received minimal early engagement with 3 points and no comments. The examples suggest a potential bridge between web development skills and native application building.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

Python's Per-Call TLS Handshake, Not Slow APIs, Caused False Latency Readings

A developer spent 48 hours investigating an apparent performance gap between curl and a Python probe hitting the same local health endpoint. While curl reused a single TCP and TLS connection across its run, the Python loop using urllib.request.urlopen() opened and closed a fresh connection on every iteration, adding handshake overhead each time. The discrepancy had nothing to do with the network, DNS, proxies, or the upstream API — all of which returned consistent HTTP 200 responses throughout. The root cause was a measurement design flaw: connection setup costs were being recorded as origin latency. The fix is to reuse connections via a persistent session object, which avoids repeated TLS handshakes and produces accurate timing data.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer builds browser extension for pixel-level image diffs in GitHub PRs

A developer has released gh-pixel-diff, a browser extension that adds explicit pixel-difference highlighting to GitHub's existing image comparison tools in pull requests. The tool was created to address the need to pinpoint exact visual changes—such as padding shifts or element movement—without manually switching between GitHub and external test reports like Playwright. It works locally in the browser with no server-side image processing, and supports GitHub, GitLab, and self-hosted instances of both platforms. Users can zoom, pan, step through changed regions, switch between multiple diff views, and export the result as a PNG for review comments. The extension is available on the Chrome Web Store and Firefox Add-ons, with Safari requiring manual installation via Xcode, and the source code is MIT-licensed.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Vue Composables Can Silently Share State Across Components Due to JS Scoping

A common but subtle bug in Vue's Composition API causes composables to share state across multiple component instances when reactive variables are declared at module scope rather than inside the function body. Because ES modules are initialized once per import, any ref or reactive value placed outside the exported function becomes a singleton shared by every caller. This means two components using the same composable can inadvertently read and mutate each other's data without any warning from Vue. The fix is straightforward: declaring state inside the exported function ensures each component call creates its own independent closure. Developers building server-rendered Vue apps face greater risk, as shared module-level state can leak data between different users' requests.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

CSS 'appearance: base-select' lets developers style native dropdowns without rebuilding them

Frontend developers have long resorted to building custom dropdown components from scratch whenever designs require icons or rich styling inside select menus, since native HTML select elements can only render plain text. These hand-built alternatives introduce complex bugs, such as popup clipping caused by ancestor elements with overflow: hidden, and require manual scroll and resize listeners to stay positioned correctly. A bigger hidden cost is accessibility: native select elements provide keyboard navigation, type-ahead, screen reader announcements, and focus management for free, while most custom dropdowns only partially replicate this behavior. A new CSS property, appearance: base-select, now allows developers to style native select elements directly, including placing real markup like flag icons inside option elements, without sacrificing any built-in browser behavior. The feature also promotes the dropdown popup to the browser's top layer, eliminating clipping issues entirely; it is currently available in Chromium-based browsers, with Firefox and Safari support still in progress.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Next.js 16 Renames middleware.ts to proxy.ts: What Teams Need to Know

Next.js 16 introduced proxy.ts as the replacement for middleware.ts, the file responsible for intercepting requests before any application code runs. The framework kept the old filename functional during the transition, causing many teams to miss the change entirely after upgrading. Unlike its predecessor, proxy.ts no longer gives developers a choice of runtime, removing the Edge-only constraint that had shaped how auth and redirect logic was written for years. Developers can migrate existing middleware.ts files using an official codemod that also handles renamed configuration options. Teams running both filenames in the same codebase after a partial migration risk inconsistent request-handling behavior that may be difficult to detect.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

CSS Units svh, lvh, and dvh Replace JavaScript Hacks for Mobile Viewport Fixes

For years, developers used a JavaScript workaround to fix the '100vh bug' on mobile browsers, where full-screen sections appeared taller than the actual visible area due to browser UI elements like address bars. The root cause is that 100vh measures the browser's largest possible viewport, not the current visible height, causing content like CTA buttons to be hidden behind toolbars on page load. iOS Safari and Chrome on Android historically handled the unit differently, making the bug inconsistent and hard to reproduce. Since 2022, all major browser engines support three new CSS viewport units — svh, lvh, and dvh — which precisely measure the small, large, and dynamic viewport heights respectively. Developers can now replace the JavaScript resize-listener hack with a single CSS declaration using dvh or svh, eliminating layout jumps and reducing code complexity.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Open-Source Legal AI Platform Mike Receives Citation Order Bug Fix

A contributor successfully merged Pull Request #466 into Mike, an open-source legal AI platform hosted under the open-legal-products organization with over 4,300 GitHub stars. The fix addressed a bug in the verifyQuoteAgainstSource function where citations spanning page breaks were incorrectly marked as verified regardless of document order. Specifically, a segment appearing after a page break could receive a verified status even when it preceded the referenced source in the actual document. The patch enforces correct document ordering across page-break segments to ensure citation accuracy. Accurate citation order is considered critical in legal research, making the fix significant for the platform's reliability.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Cuts CrimeLens API Latency by 49x With Caching, Indexing and Redis

A developer has completed a performance optimization cycle for CrimeLens, a full-stack crime reporting and analytics platform built with React, Node.js, PostgreSQL, and PostGIS. Before optimization, the system showed severe latency under load, with p95 response times reaching 5.13 seconds under normal traffic and 36.64 seconds during stress tests. Improvements included adding PostgreSQL indexes, Redis caching, rate limiting, response compression, Nginx reverse proxying, and containerization via Docker. After these changes, median latency dropped from 1,230 ms to just 25 ms, while stress-test throughput jumped from 32.3 to 282.2 requests per second. The HTTP error rate also fell roughly 16.7 times, and the system handled over eight times more requests under stress without introducing new errors.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Releases BHive, a TypeScript Multi-Agent AI Framework, Seeks Peer Review

A developer has publicly released BHive, a TypeScript-based multi-agent framework built over two weeks, and is soliciting critical technical feedback before committing further resources. The framework supports multiple LLM providers — including Gemini, OpenAI, Anthropic, and local Ollama — through a unified LLMAdapter interface. Individual agents, called 'Bees,' can delegate tasks to one another, call tools, and operate via a shared provider abstraction layer. BHive includes a security model with injection guards, allowlists, and size limits, and currently ships with 299 passing tests at roughly 98% code coverage. The developer is specifically seeking scrutiny on architectural decisions such as tool-call validation placement, timeout configuration depth, and the delegation pattern design.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Google Business Profile Adds Transparency Tab Showing Automated Data Collection

Google is rolling out a new 'Collected Info' tab in Google Business Profile dashboards, giving eligible business owners visibility into data its automated assistant has gathered about their company. The tab displays each collected item along with its source and the date it was collected, and allows owners to delete individual entries. Data may originate from automated phone, text, or WhatsApp interactions, as well as user contributions, public reports, and licensed content. Importantly, deleting a collected entry removes it from Google's records but does not automatically reverse any changes already applied to the public listing. The feature is being introduced in phases and is currently limited to select regions, languages, and business categories.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Study finds most high-comment GitHub repos have zero outside contributor engagement

An analysis of GitHub repositories found that seven repos appeared to comment on 40–80% of their issues, but deeper sampling revealed a starkly different picture. After filtering out pull requests and applying a minimum 30-issue threshold, five of the seven repositories showed a third-party commenter rate of exactly zero across a 25-issue sample. This suggests that most of the commenting activity in these repos comes solely from the issue authors themselves, with no external community participation. The researcher also surveyed 21 similar repositories active since August 2026 and found none that met both a high comment rate and meaningful third-party engagement. The findings indicate that high issue-comment counts can be misleading, often reflecting self-contained notes rather than genuine community conversation.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

BullSaddle Offers Developers a Read-Only Dashboard for Managing Git Workspaces

BullSaddle is a local-first developer tool designed to simplify the management of multiple Git repositories spread across a machine. It provides an operational overview of the entire Git workspace, including repositories, worktrees, diagnostics, and repository health metrics. The tool is strictly read-only, meaning it does not modify or interfere with existing Git repositories. Its core focus is on making complex, multi-repository workspaces more observable and easier to monitor. BullSaddle is available as an open-source project on GitHub, developed by Orion Impact.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Why Compound Interest Code Always Breaks — and How to Debug It

Finance features in software products frequently produce results that diverge from spreadsheet calculations, but the error is rarely in the formula itself. The root cause is almost always a modelling assumption — such as misinterpreting APY versus APR, misaligning time units, or incorrectly placing when contributions land. A developer writing on DEV Community outlines a checklist-based debugging approach for compound interest implementations, drawing on repeated real-world experience with savings and projection tools. The guide covers five hidden assumptions in the standard formula A = P(1 + r/n)^(nt) and explains how each can silently corrupt results over long time horizons. It recommends reproducing bugs with minimal inputs, comparing output against a trusted external calculator, and logging every derived value to isolate discrepancies efficiently.

0 comments Read more at DEV Community
← NewerPage 1010 of 3192Older →