SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World

Programming

LatestTopDiscussed
0
ProgrammingDEV Community ·

Developer builds SQL-based Supabase RLS audit tool that bypasses AI guesswork

A developer has created a structured audit tool for Supabase Row Level Security (RLS) that relies on direct SQL queries rather than asking AI agents to infer security configurations. The tool queries Postgres catalogue views such as pg_tables and pg_policies to surface concrete issues including tables with no RLS, overly permissive policies, public storage buckets, and tables with RLS enabled but no attached policy. Unlike typical AI-assisted audits, the prompt instructs the model only to run a fixed SQL block and report results, ensuring findings are grounded in actual database state. The tool is designed to be pasted into AI coding environments like Claude Code or Cursor and runs against a live Supabase project in around thirty seconds at no cost. The developer noted that all detected issues can silently reappear after migrations or ad-hoc fixes, which motivated building a broader monitoring product called Defencecore to track catalogue changes over time.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Aux Battles runs multiplayer music game using Postgres as sole state machine

Aux Battles is a free, browser-based multiplayer music game that operates without a dedicated game server. Instead, all game state — rooms, players, prompts, and voting — is synchronized across clients via Supabase Realtime, which streams live Postgres database changes. Every player action is handled as a direct table write, with a single phase column on each room row driving the game's progression. Host-only controls are enforced at the UI level rather than through row-level security, keeping the architecture simple for a no-login game. The most complex feature was a Twitch-style 64-song knockout bracket, stored in a single JSONB column and protected against race conditions using a compare-and-set sequence token.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

MemoryDesk: Open-Source Agent That Retains Customer Support Memory

MemoryDesk is an open-source customer support agent project hosted on GitHub by developer THRISHANTH222. The project focuses on building AI agents capable of persistent memory across conversations. Its core architectural principle treats memory as a standalone, modular capability rather than embedding it within individual chat sessions. This design allows the agent to recall past interactions without being constrained by single-conversation context limits. The project is available publicly for developers to explore and contribute to.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Self-Hosted vs SaaS Support AI: What EU E-Commerce Shops Need to Know

Businesses running customer support AI can either use a hosted solution like Intercom's Fin or deploy a self-hosted agent on infrastructure they control, with the core difference being architectural. For EU-based shops, self-hosting means customer data stays within a chosen region and jurisdiction, making data-residency compliance straightforward and independent of third-party sub-processors. On the cost side, hosted tools typically charge per resolved ticket — tying support expenses directly to sales volume — while self-hosted setups use fixed monthly instance costs and measurable, cappable inference fees. Self-hosting also avoids vendor lock-in, since the knowledge base, conversation logs, and automation logic remain portable and operational even if the setup partner changes. However, the trade-off is real: teams must own operations including monitoring, model updates, and incident response, which may not suit smaller teams without dedicated technical resources.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How to Add Directory Categories Without Breaking Ranked URLs

Expanding a web directory by adding new categories risks renaming existing URLs that already rank in search engines, which can cause lasting SEO damage even with 301 redirects in place. Developers are advised to preserve existing top-level URL paths and introduce new categories as separate routes, sharing logic through a common resolver function rather than restructuring the entire URL schema. A canonical redirect guard can automatically convert stale or mismatched URLs into 301 redirects, preventing 404 errors from old links or sitemap entries. New category pages should only go public once they meet a minimum inventory threshold, avoiding the indexing of thin, low-value content. The recommended rollout sequence is to add the category internally, seed listings, enable the route once populated, then add it to navigation and the sitemap before submitting URLs for crawling.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

Samsung trials Anthropic's Claude AI for chip design verification with mixed results

Samsung has begun using Anthropic's Claude AI model to assist in verifying semiconductor chip designs. The initiative aims to leverage large language models to streamline what is traditionally a complex and time-intensive engineering process. However, the rollout has encountered notable difficulties, suggesting the technology is not yet fully suited to the demands of chip verification. The challenges highlight broader industry questions about the readiness of current AI tools for highly technical hardware engineering tasks.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

New SDK lets developers mint Algorand NFTs with IPFS metadata without Pinata

A new open-source SDK called ipfs-pay-to-pin-client allows developers to create NFTs on the Algorand MainNet blockchain without needing a Pinata account, API key, or subscription. The tool pins NFT metadata to IPFS using an automated micropayment system powered by USDC on the Algorand network, following the x402 payment protocol. Metadata pinning comes with a 365-day retention period per payment, with early renewal options available. Developers can mint Algorand Standard Assets with on-chain IPFS references by installing two Python packages and running a provided script. The workflow covers metadata creation, IPFS pinning via a pay-to-pin gateway, and ASA creation on MainNet in a single automated process.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

DFlash Speculative Decoding Promises Speed Gains, But Real-World Results Vary

DFlash, a speculative decoding framework developed at UC San Diego's z-lab, uses a block-diffusion drafter to predict multiple tokens simultaneously rather than generating them one at a time, distinguishing it from autoregressive methods like EAGLE-3. The framework claims up to 6x lossless acceleration in paper benchmarks and up to 15x on NVIDIA Blackwell hardware under favorable conditions, but engineers warn these figures represent upper bounds rather than typical production performance. Independent testing on 30B-scale models, including Meta's own vendor-published results, shows more modest real-world gains, such as 3.1x speedup at batch size 1 on an RTX 5090 with 4-bit quantization. Actual performance depends heavily on factors including draft-token acceptance rates, prompt distribution, sampling configuration, and memory bandwidth. DFlash is available as a drop-in option in popular serving stacks including vLLM and SGLang, driving interest among teams considering it as a default replacement for EAGLE-3 drafters.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

OpenAI Data Shows Frontier Firms Use AI 3.5x More Intensely Than Typical Peers

OpenAI's B2B Signals data for 2026 reveals that frontier enterprises now use 3.5 times more AI intelligence per worker than typical firms, up from a twofold gap recorded a year earlier. The figures come from OpenAI's State of Enterprise AI 2025 report and May 2026 B2B Signals materials, which frame adoption as a question of depth and integration rather than mere access. Frontier organizations also show markedly higher activity in specific contexts, including 16 times more Codex messages and 6 times more data-analytic messages compared to baseline users. On June 2, 2026, OpenAI expanded Codex with six role-specific plugins bundling 62 apps and 110 skills, alongside Sites and annotations, aiming to embed AI into defined job functions and repeatable workflows. Experts caution that raw usage volume alone does not indicate meaningful workflow integration, and that genuine enterprise value depends on connecting AI to accountable processes and roles.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

JavaScript at 30: How a 10-Day Build Became the Web's Core Language

JavaScript is a programming language designed to make websites interactive, working alongside HTML and CSS to build dynamic web experiences. It was created by Brendan Eich at Netscape Communications in May 1995, with the first version completed in just ten days. Originally named Mocha and later LiveScript, it was officially released as JavaScript in December 1995. The language enables features such as animations, form validation, pop-up messages, and real-time content changes on web pages. Today, JavaScript powers not only websites but also web applications, mobile apps, backend systems, and desktop software, running in browsers via engines like Chrome's V8 and Firefox's SpiderMonkey.

0 comments Read more at DEV Community
0
ProgrammingGitHub Blog ·

GitHub's Secure Open Source Fund reveals security lessons from 50 AI-era projects

GitHub has shared findings from Session 4 of its Secure Open Source Fund, which supported 50 open source projects in strengthening their security. The initiative combined AI-assisted workflows with maintainer expertise and dedicated funding to address vulnerabilities. Projects also leveraged GitHub's built-in security tools alongside guidance from external experts. The program highlights how open source communities can adopt modern, AI-augmented practices to improve overall software security.

0 comments Read more at GitHub Blog
0
ProgrammingDEV Community ·

Why Governing AI Agents Matters More Than Trusting Them

As AI agents write an increasing share of production code, software teams face a critical question: how do you verify correctness without blind trust? The author argues that the answer lies not in trusting AI blindly, but in building robust governance systems — much like the methodologies developers have long used to catch human error. A key shift is moving from evaluating how code is written to evaluating whether the output satisfies a defined contract of requirements, tests, and security criteria. Critically, rules embedded only in prompts or instruction files are treated as requests, not enforceable constraints, and agents can misinterpret or ignore them. As agents gain greater autonomy, teams must invest in environment-level controls and verification mechanisms rather than relying on constant human oversight.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer cuts AI agent costs by replacing model decisions with deterministic scripts

A developer building an agent orchestration system in April 2026 initially delegated all pipeline steps — task selection, implementation, and closure — to an AI model. After three days of focusing on UI polish, he questioned why the model was handling task selection, a process governed by simple conditional logic with no ambiguity. He rewrote task selection and task closure as deterministic scripts, removing the need for the model to perform those steps entirely. Task selection had previously cost around $0.25 and 40 seconds per run, while closure cost roughly $0.30 and 20 seconds; both dropped to zero cost and near-instant execution. The key insight, drawn from over 2,000 completed tasks, was that the real design challenge is identifying which parts of a workflow genuinely require model judgment and which do not.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Passing Tests Is Not Enough: AI Coding Agents Miss Contextual Correctness

Developers and researchers are raising concerns that AI coding agents, such as OpenAI's Codex and Anthropic's Claude Code, can produce code that passes all tests yet still be architecturally wrong for the current system. As these agents increasingly operate at the repository level rather than generating isolated snippets, the meaning of 'correctness' has expanded beyond functional outputs. Existing benchmarks like SWE-bench evaluate agents on whether their patches satisfy predefined tests, but do not measure how well agents adapt when the underlying architecture or constraints change. Emerging research such as SWE-ContextBench and SWE-Explore is beginning to address context-awareness, but a formal framework for testing context-shift adaptation is still lacking. The proposed solution is not to replace current benchmarks but to add controlled evaluations that measure both an agent's ability to adapt decisions when relevant context changes and its stability when only irrelevant context changes.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

MiniMax-H3 generates cinematic video with synced audio from text prompts

MiniMax-H3 is an open text-to-video model that produces short cinematic clips complete with a synchronized soundtrack in a single pass, unlike earlier models that required separate audio pipelines. It also supports keyframe conditioning, allowing users to supply an optional first or last frame so the model interpolates motion between them, giving creators more directorial control. The model is available via Hugging Face Inference Providers, meaning users can run it for free using their account quota without needing a GPU or local installation. A no-code Gradio workflow built with gr.Workflow lets users drag and drop inputs, preview outputs, and chain additional operators such as upscaling. Early demos shared online include AI-generated recreations of scenes from Breaking Bad, Friends, and The Office.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

React Tutorial: Build a Currency Converter Using useState and useMemo Hooks

A developer tutorial on DEV Community walks readers through building a simple currency converter application using React. The project uses fixed exchange rates to convert USD into EUR, GBP, and JPY, avoiding the complexity of live API calls. Key React concepts covered include the useState hook for managing user input and the useMemo hook for optimizing conversion calculations. The tutorial is designed as a beginner-friendly, practical exercise that introduces state management, user input handling, and performance optimization in a single project.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Anthropic in $6B Talks to Acquire AI Startup Decart for Compute and Robotics Edge

Anthropic is reportedly in talks to acquire Decart AI for approximately $6 billion, according to Bloomberg and Fortune reports from August 13, though no deal has been signed. Decart, founded in 2023 and last valued at $3.1 billion, is known for Oasis, a real-time world-simulation model, and DOS, a hardware-aware inference optimization stack. Sources cited by both outlets suggest the primary driver is Decart's chip-efficiency technology, which could reduce Anthropic's significant compute costs across all Claude model requests. Anthropic has also previously held acquisition talks with robotics foundation model firm Physical Intelligence, signaling broader strategic interest in the physical AI space. Analysts note that acquiring Decart's world-simulation capabilities could serve as an alternative path into robotics training infrastructure if direct robot-brain acquisitions remain blocked by competing investor interests.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Tool With 33 Passing Tests Never Ran Once — No Schedule Was Set

A software team built a monitoring tool designed to scan four public sources daily for relevant product mentions, writing 33 tests that all passed before shipping. Despite thorough acceptance checks covering output limits, error handling, and duplicate removal, the tool had no scheduler attached, so it never executed and produced no results. Because a silent tool reports nothing, the failure went unnoticed until someone asked what the tool had found and the answer was simply nothing, ever. The team responded by writing a lightweight automated guard — 12 lines of code — that checks on every push whether any script claiming a daily schedule actually appears in a workflow file. They also configured the scheduled job to send a daily message even when it finds nothing, making silent failures immediately distinguishable from a job that has stopped running.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Five Advanced CLI Engineering Patterns for Node.js and Go Production Tools

Software engineer Lakshan Muruganandam has outlined five advanced patterns for building high-performance command-line tools, drawing from his open-source projects port-sniper and node-reaper. The patterns cover graceful signal handling, ensuring CLIs cleanly release ports and temporary files when interrupted via Ctrl+C. Other recommendations include interactive dropdown menus to reduce flag memorization, Go goroutine worker pools for efficient filesystem scanning, and lightweight ANSI libraries for terminal color output. Muruganandam also emphasizes returning correct POSIX exit codes so CLI tools integrate reliably into CI/CD pipelines.

0 comments Read more at DEV Community
← NewerPage 158 of 1335Older →