SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World

Programming

LatestTopDiscussed
0
ProgrammingDEV Community ·

Google shifts TPU management to Compute Engine as Cloud TPU API enters maintenance mode

Google has stopped active development of the Cloud TPU API, limiting it to bug fixes and security patches only, with no published sunset date. New TPU hardware generations, beginning with TPU7x (Ironwood), are exclusively supported through Compute Engine or Google Kubernetes Engine. A developer documented migrating a v6e-1 (Trillium) chip running the Gemma model under vLLM from the old queued-resources API to the new Compute Engine instances workflow. The core change involves replacing TPU-specific CLI flags and commands with Compute Engine equivalents, such as swapping accelerator-type for machine-type and using gcloud compute instances instead of gcloud compute tpus. Developers on v5p and v6e can migrate without special access, while TPU7x flex-start provisioning currently requires allowlist approval from Google's account team.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

FC Bayern Signs Long-Term AI Partnership With Google, Names Gemini Official Partner

FC Bayern Munich and Google have announced a long-term partnership that designates Gemini as the German football club's official AI partner. The deal aligns one of Europe's most prominent football clubs with Google's generative AI brand, though no specific products, features, or rollout timelines have been disclosed. It remains unclear how Gemini will be integrated across fan-facing channels, internal operations, or commercial services. The partnership is currently a strategic and branding agreement, with implementation details yet to be defined. Further announcements are expected to clarify whether the collaboration results in visible supporter tools or primarily serves as a commercial affiliation.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

Nvidia Launches Nemotron 3.5 Lightning and Nemo Switchyard AI Tools

Nvidia has announced two new AI offerings: Nemotron 3.5 Lightning and Nemo Switchyard, detailed on the company's official blog. The tools appear to be designed for deployment across Nvidia's hardware ecosystem, including RTX and DGX platforms. The announcement was shared on Hacker News, though it attracted minimal community discussion at the time of posting. Further technical details are available via Nvidia's blog, which outlines the capabilities and intended use cases for both products.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

Developer Tests Kiro Crew AI Agent on Decade-Old Laptop as a DIY SRE Tool

A developer deployed Amazon's Kiro Crew AI agent on a aging Positivo DUO ZR3630 laptop running an Intel Celeron with 4GB RAM and 32GB storage to act as an automated SRE and personal finance assistant. The setup, running Linux Mint 22 with Python and MySQL in Docker containers, was accessed remotely via Tailscale. Kiro Crew was tasked with monitoring container health, recovering failed services, and sending a daily Telegram report of credit card spending pulled from a MySQL database. In testing, the agent successfully detected and recovered from simulated container failures, including a downed backend and a broken MySQL service. Even after crashing during a memory exhaustion test, Kiro Crew recovered and continued functioning, surprising the developer with its resilience on such constrained hardware.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Top AI Frontend Code Generators Compared: v0, Make Your UI, Lovable and More

AI-powered frontend generators have reshaped how developers build web interfaces, allowing them to produce UI code from text prompts, screenshots, or design files in seconds. A 2026 comparison of leading tools — including v0 by Vercel, Make Your UI, Lovable, Bolt.new, Replit, and Framer AI — highlights key differences in scope and specialization. v0 excels at React and Next.js component generation, while Make Your UI stands out for supporting multiple frontend stacks such as Vue, Angular, and Svelte without touching backend logic. Some tools like Lovable and Bolt.new extend beyond the frontend to generate full-stack applications including databases and authentication. Developers with existing backends are advised to consider dedicated frontend generators to retain control over their architecture rather than using broader app-building tools.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Why Go developers rejected syntactic sugar and kept verbose error handling

A Medium post published on August 9, 2026, revisiting a Go versus Rust backend rewrite reignited debate over Go's explicit error handling pattern, the widely repeated 'if err != nil' check. Despite 13% of respondents in the April 2024 Go Developer Survey citing error handling verbosity as a challenge, the language maintains a 93% satisfaction rate among developers. In June 2019, Go co-designer Robert Griesemer proposed a built-in 'try()' function to reduce boilerplate, but the community rejected it overwhelmingly and the proposal was closed in July 2019. Go's philosophy, articulated by Rob Pike in 2015, treats errors as programmable values rather than hidden control-flow mechanisms, forcing developers to explicitly decide how each failure is handled. Major companies including Uber, Cloudflare, and Monzo rely on this pattern in production, using Go 1.13's error-wrapping tools to produce readable, traceable error chains.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Typosquatted Ruby Gem Nearly Merged, Exposing Gaps in Supply Chain Security

A developer team narrowly avoided installing a malicious Ruby gem after a reviewer spotted a one-character typo in its name during a pull request — a catch attributed to luck rather than any formal process. The gem's extconf.rb file was designed to execute arbitrary code at install time, meaning damage could occur on developer machines or CI runners before the package ever reached production. The incident highlighted over-reliance on RubyGems.org's automated defenses, which the registry itself estimates catch only 70–80% of malicious packages, leaving a significant residual risk. Security experts recommend a layered defense approach including pinning and committing Gemfile.lock, avoiding floating version ranges, and enabling Bundler 2.6's checksum verification to detect tampered packages. The case underscores that supply chain threats such as typosquatting, lockfile injection, and post-disclosure vulnerabilities are routine background noise in modern software development, not rare edge cases.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

BlueMove DEX Loses 714,000 SUI to Cross-Version Reserve Desync Exploit

On July 11, 2026, an attacker drained approximately 714,000 SUI (around $528,000) from BlueMove DEX, a Sui-based automated market maker, within just 23 minutes. The stolen funds were quickly moved across a Wormhole bridge as USDC, and because the contracts were immutable, there was no way to freeze or patch them. BlueMove attributed the incident to an arithmetic overflow bug in a legacy contract, while a researcher at Quantum Void Labs suggested a suspicious function introduced in a May 31 upgrade played a role. An independent technical analysis found the actual cause to be a reserve desynchronization between two callable versions of the same package, where V1 and the latest version wrote different values to the reserve variable. This discrepancy allowed attackers to deflate the recorded reserve, inflate LP token minting ratios, and systematically empty the liquidity pools.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Study finds AI agents with memory-first design adopt 100% of false facts

A developer ran a controlled experiment on a Python codebase intelligence server to test how AI agents handle poisoned memory containing a mix of true and false facts. Three agent configurations were tested: a no-memory baseline, a code-first honest agent, and a memory-first lazy agent. The memory-first agent, which mirrors many token-optimizing production systems, blindly trusted its memory store and adopted 100% of the injected false facts without verifying against the actual code. Even the more cautious code-first agent showed a 12% false-fact adoption rate, specifically for plausible claims about external systems that the codebase neither confirmed nor denied. A follow-up experiment implementing a 'Verify-On-Read' mechanism was reported to have closed that remaining 12% contamination gap.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

18-Year-Old Developer Builds and Deploys Two Apps Entirely on a Mobile Phone

An 18-year-old developer has built and launched two web projects — a Python learning app and a racing game called Nitro Rush Revamped — using only an Android smartphone. The tools used include Replit, Lovable, Spck Editor, and GitHub Mobile, with some AI assistance for development. Nitro Rush Revamped features multiple vehicles, joystick controls, a leaderboard, missions, and an end-game stats dashboard. The developer is about to enter engineering college and created both projects entirely on a phone screen without access to a desktop. Both projects are live and available on GitHub as open-source repositories.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Laravel APP_DEBUG=true in Production Exposed Source Code, SQL Queries for Months

A development team discovered last week that their Laravel application had been running with APP_DEBUG set to true in a production environment for several months. The misconfiguration caused every crashing route to display Laravel's full debug screen publicly, exposing file paths, stack traces, source code, SQL queries, and in some cases environment variables. The issue came to light during a routine SEO audit, when a Google Search Console alert flagged 2,246 pages returning 5xx server errors. Google had indexed all these error pages, meaning the sensitive technical details were accessible to anyone who found them via search. The team has since documented how they identified the breach, assessed the damage, and built a pre-deployment checklist to prevent similar oversights.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Software Engineer Pivots From SDE Track to Build Subsea ROV Portfolio

A software engineer with eight years of professional experience has decided to shift career focus away from traditional SDE roles, citing concerns about the industry's trajectory. Since early 2026, work on robotics integration at Nio-Robotics prompted a pivot toward subsea robotic exploration and autonomous underwater vehicles. The engineer aims to work as an ROV engineer or survey specialist aboard research vessels in the Pacific or Atlantic Ocean. To build a credible portfolio, they are setting up a reproducible DevContainer environment to simulate an ROV using ROS2 Jazzy and Gazebo, targeting platforms like the BlueROV2. The goal is to create a publicly accessible GitHub repository that anyone can launch, serving as a foundation for real-world subsea deployment work.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

OpenAI's ChatGPT Desktop App Officially Supports Windows and macOS, Not Linux

OpenAI's official documentation for the ChatGPT desktop app covers only Windows and macOS, with no Linux client, preview, or installation path announced. Third-party Linux wrappers and community ports exist but are independent projects and do not carry OpenAI's support, security guarantees, or release process. A past macOS security update noted the issue did not affect Linux, but this refers solely to that incident and should not be read as confirmation of a Linux app. Organizations evaluating desktop AI tools should verify whether any client is first-party software backed by official documentation before deploying it on managed systems. Until OpenAI formally announces Linux support, Linux remains outside the confirmed ChatGPT desktop platform footprint.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

Developer Creates Single-Page Visual Map of the Entire PyTorch Framework

A developer has published a comprehensive single-page reference guide covering the entirety of the PyTorch machine learning framework. The resource is hosted at tensor.khalilli.ai and aims to serve as a structured map for navigating PyTorch's concepts and components. The post was shared on Hacker News, where it received 8 upvotes. The guide appears intended to help developers and researchers quickly orient themselves within the PyTorch ecosystem without consulting scattered documentation.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

Developer finds only 1 of 591 pages indexed by Google, shares diagnostic steps

A developer running a directory site with 2,499 vendor listings across six Middle Eastern and North African cities discovered that Google had indexed just one of 591 programmatic landing pages. After verifying that robots.txt, sitemaps, internal linking, structured data, and server-rendered content were all correctly configured, the root cause remained unclear. The key diagnostic step was checking Google Search Console's indexing report, which distinguishes between pages that are 'Discovered but not indexed' and pages that are 'Crawled but not indexed' — two statuses that require opposite fixes. The latter status typically signals a content quality rejection, meaning building backlinks would have no effect and the real solution is consolidating pages with more unique, substantive content. The developer cautioned that programmatic SEO can create a false sense of progress, as generating hundreds of thin pages may not improve indexing if Google has already deemed the existing ones low quality.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

WooCommerce default setting silently inflates checkout prices for EU digital product buyers

A self-hosted WooCommerce store owner discovered that a digital product listed at €49 was charging Hungarian customers €51.85 at checkout — a discrepancy caused by EU VAT rules requiring tax to be applied at the buyer's local rate. By default, WooCommerce stores the entered price as a pre-tax figure and recalculates the final amount per country, meaning the displayed price and checkout price can diverge without any error or warning. The issue stems from a fundamental trade-off: either the consumer-facing price stays fixed across the EU while the seller's margin varies, or the margin stays fixed while the price changes per country. WooCommerce silently defaults to protecting the seller's margin, which can violate EU consumer pricing rules that require the advertised price to match what customers actually pay. The problem can be resolved with a single line of code — a WooCommerce filter that locks the tax-inclusive price across all regions — but the setting is not enabled by default.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

5 Reusable AI Prompts a Freelance Developer Uses to Save 10+ Hours Weekly

A freelance developer with three years of AI-assisted experience has shared five specific, reusable prompts designed to cut 10–15 hours from a typical workweek. The prompts address common developer pain points including debugging, converting vague client requests into structured tickets, and understanding legacy code. One prompt instructs AI to act as a senior engineer asking clarifying questions rather than immediately offering fixes, helping surface bugs through structured reasoning. Another extracts meeting transcripts into a formatted table of decisions, open questions, and action items with owners. The author argues that consistent reuse of a small set of targeted prompts delivers more practical time savings than chasing the latest AI models or tools.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

LLM eval tools offer 50+ metrics, but the hard parts of evaluation remain unsolved

A developer spent a week reviewing the metric catalogs of five widely-used LLM evaluation libraries — Arize Phoenix, DeepEval, Future AGI, Langfuse, and Ragas — to assess what they actually offer beyond their marketing claims. All five tools provide ready-made metrics, with counts ranging from around 16 to 72, and every one supports LLM-as-a-judge scoring. However, the author argues that the metrics themselves represent only the easier 20 percent of the evaluation problem. The genuinely difficult tasks — selecting a metric that matches a specific failure mode and quantifying uncertainty in results — are left entirely to the user. Notably, the catalogs are converging on the same core metrics, such as Faithfulness and Tool Correctness, suggesting that metric coverage is fast becoming a commodity rather than a differentiator.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Builds Browser-Based Bulk HTML-to-Markdown Converter Using Go and WebAssembly

A developer has released a free tool called HTML to Markdown AI that converts large batches of HTML files into GitHub-Flavored Markdown entirely within the browser. The tool was created to address a personal need: converting hundreds of mirrored website pages into clean Markdown for use in LLM knowledge bases, without relying on any server-side processing. Users can drop a ZIP archive or individual HTML files into the interface, and a Go-based conversion engine compiled to WebAssembly handles all processing locally, meaning files never leave the user's machine. The pipeline strips boilerplate elements such as navigation and scripts, extracts main content, and preserves folder structure in the output ZIP. The tool is built with Astro and Tailwind on the frontend and is aimed at developers working on RAG pipelines, documentation exports, and AI training data preparation.

0 comments Read more at DEV Community
← NewerPage 211 of 1341Older →