SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World

Programming

LatestTopDiscussed
0
ProgrammingDEV Community ·

How to Poll Transactional Email Status in Node.js Using Cron and Events API

Developers can track transactional email delivery status in Node.js by polling a provider's events API via a cron worker, without relying on webhooks. The approach involves recording an immutable send attempt with an internal ID, provider message ID, template revision, and a digest of the rendered body for auditability. A bounded polling strategy — checking once per minute for up to ten minutes — is recommended, after which observation stops regardless of delivery outcome. Provider status labels like 'delivered' or 'bounced' reflect transport state only and should never influence security logic such as password-reset token expiry, which must remain an independent policy. This polling method is not suitable for scenarios requiring sub-minute reaction times or high provider-side event volumes, where a webhook-based push approach is more appropriate.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Stanford's Paper2Agent Converts Research Repos into Executable AI Assistants

A Stanford University research team has developed Paper2Agent, a system that transforms academic research repositories on GitHub into interactive AI agents capable of running actual code. The tool clones a research repository, executes tutorial files, extracts functional components, and assembles them into a standardized tool server that users can query directly. Published in Nature on September 16, the system was tested on 100 bioinformatics papers from bioRxiv, successfully converting 74, while achieving roughly 98% accuracy on a separate set of 10 non-biology repositories. Failures were attributed to incomplete code, poor documentation, or unresolvable dependencies — limitations the team disclosed transparently. The project is open-source under an MIT license but requires an external AI agent such as Claude or Gemini CLI to operate, with API costs estimated at around $14–15 per complex repository and processing times ranging from 45 minutes to three hours.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Dev Shares 20-Minute Daily Routine That Keeps His SaaS Business Running

A developer building FoxyInvoice in public has detailed the daily operating routine he credits with sustaining the business, breaking it into four steps totalling around 20 minutes. The process begins with a five-minute triage of three inboxes — user feedback, a lead queue, and app error reports — followed by a ten-minute automated scan of Hacker News and Reddit for relevant problem-language mentions. Machines handle the scanning and pre-scoring, while the founder reserves the final five minutes for human judgment on whether and how to respond to flagged conversations. A fourth step feeds recurring user questions back into the product to drive improvements over time. The author argues that consistent daily habits, rather than new features, are what ultimately keep a software product alive.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

In the AI Age, Knowing What You Want Matters More Than Ever

As AI tools increasingly handle tasks like writing, decision-making, and content recommendations, many people still report feeling directionless despite having more guidance than any previous generation. The core issue, according to a DEV Community analysis, is not a lack of information or tools but an absence of personal clarity about goals and values. While AI excels at execution — coding, research, automation — it cannot determine which objectives are actually meaningful to a given individual. Recommendation algorithms optimize for engagement, not personal fulfilment, meaning they can efficiently direct attention toward things that do not align with a user's deeper priorities. The article argues that a strong internal compass — a clear sense of values, principles, and direction — is the foundational layer that makes powerful technology genuinely useful rather than a faster path to the wrong destination.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How one developer split a 170 GB Rust project across repos to fix CI gridlock

A developer running a self-hosted CI runner found that a single Rust project was consuming over 170 gigabytes of build artefacts per run, blocking all other repositories for up to an hour at a time. The machine's disk had to be manually cleared after each build cycle, and every unrelated project — from Go modules to a blog — queued behind the Rust crate graph. While Cargo workspaces offered a built-in solution for incremental compilation, the developer found they couldn't provide independent release cadences for reusable library crates. The solution was a hybrid approach: keeping a workspace for tightly coupled distributable components while moving reusable libraries like config, TUI, and telemetry into separate repositories. This allowed eleven crates to maintain five distinct version numbers, so unchanged crates like tui were no longer dragged into unnecessary releases.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Same mutation score, different denominators: why CI and local results silently diverged

A developer noticed that their mutation score of 1.000 appeared in two places but with different denominators — 168/168 in CI and 170/170 locally — despite the score looking identical. The discrepancy arose because the local working tree had 28 uncommitted files, including modified test suites and fixtures, which changed which mutants were caught and which refusal sites were excluded. This effectively meant the local run measured a different system rather than the same one at a different quality level. The developer had even written a CI comment months earlier explicitly warning that mutation scores are only comparable when run against the same suite. The incident highlights how a perfect headline metric can mask underlying inconsistencies, and raises the question of whether published repository metrics are truly stable across a clean clone versus a local working tree.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Why Monitoring and Telemetry Matter After You Launch an AI-Built App

AI tools have made it easier than ever for beginners to build and ship functional apps, but keeping those apps running smoothly requires visibility into real-world usage. Monitoring and telemetry — which include logs, error reports, and performance data — help developers understand what their app is actually doing once other users interact with it. Without this data, diagnosing bugs is guesswork; with it, developers can give AI assistants specific, evidence-based context to identify and fix problems more accurately. A practical debugging workflow involves identifying an error or slow request, reviewing available logs, and then prompting the AI to explain the cause and apply a targeted fix. Croft, a development platform, has introduced built-in monitoring and telemetry for all workspace apps, requiring no additional setup or code changes.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Tencent's Gander AI Splits Brain Architecture to Handle Voice Chat and Background Tasks Simultaneously

Tencent's Hunyuan Speech Team released an open-weight AI research project called Gander in early September, designed to hold real-time voice conversations while executing background tasks concurrently. The system uses a dual-component architecture inspired by the human brain: a 'Cerebellum' module handles live voice and visual input, while a 'Brain' module processes heavier tasks like web searches or code execution in the background. Model weights are publicly available on Hugging Face under an Apache 2.0 license, built on top of the open MiniCPM-o 4.5 base model. However, running the full system requires three separate GPUs and approximately 19 GB of VRAM in full precision, making it impractical for most general users at this stage. In benchmark tests, Gander achieved around 40% task-completion accuracy on overlapping-speech evaluations, compared to roughly 60% for closed systems like GPT-Realtime, positioning it as a research milestone rather than a consumer-ready product.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Builds Battery-Efficient Android App to Auto-Silence Phone by Location

A developer created an Android app called Muffle that automatically adjusts phone ringer modes based on the user's physical location, eliminating the need for manual volume toggling. Instead of polling GPS at fixed intervals — a known battery drain — the app leverages Android's GeofencingClient API, which offloads location monitoring to the OS at the hardware level. The app runs as a ForegroundService to prevent the system from killing its background process, ensuring reliable operation even when the app has not been opened for days. To address GPS signal drift near geofence boundaries, the developer implemented hysteresis logic that verifies a user's distance from a zone's center before triggering any ringer change. The project highlights a broader challenge in mobile development: balancing location responsiveness with energy efficiency on Android devices.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How to manually set up shadcn/ui with Vite, React, TypeScript, and Tailwind CSS v4

A developer guide published on DEV Community walks through manually configuring a new project using Vite, React, TypeScript, Tailwind CSS v4, and shadcn/ui with the pnpm package manager. The tutorial is aimed at newer developers and focuses on explaining the path alias configuration that shadcn/ui requires to function correctly. Steps covered include scaffolding the Vite app, installing Tailwind via its Vite plugin, updating TypeScript config files with the necessary path aliases, and initializing shadcn/ui. The guide also adds a sample Button component to verify the full setup is working. The author notes that an automated alternative exists in the official shadcn/ui documentation for those who prefer it.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

CIDS Cybersecurity Platform Rebrands After Logo Found Similar to Stock Design

CIDS, a cybersecurity platform, has unveiled a new logo following a brand review that revealed its previous design closely resembled an existing stock image. To avoid potential intellectual property risks, the creator opted for a proactive redesign rather than continuing with the old identity. The decision was driven by a long-term vision for CIDS as a lasting cybersecurity platform requiring a solid, original brand foundation. The updated logo retains the platform's core mission, summarized by its tagline: 'Security that thinks.'

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer builds CLI tool to grade open-source projects using git history

A developer named Kenji Rasmussen, who identifies as an autonomous AI agent, created a command-line tool called gitfault that analyzes raw git logs to assess codebase health. The tool measures hotspots, change coupling, and knowledge risk — including bus factor — without requiring language plugins or machine learning. Rasmussen applied gitfault to six well-known open-source projects including bat, Flask, fzf, Express, requests, and click. Among the findings, sharkdp/bat scored highest for overall health, while junegunn/fzf's src/terminal.go stood out with 758 revisions and roughly 22,000 lines of churn. The tool is publicly available and can be installed via pipx, uvx, or Homebrew.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

38% of analyst queries return empty results due to hidden permission failures in text-to-SQL

A structural analysis of text-to-SQL systems has found that analysts receive silent, misleading 'no records found' responses 38.5% of the time when the data exists but is restricted by row-level security. The core problem is that most text-to-SQL pipelines expose the full database schema to the language model, which then generates valid SQL against tables the user cannot access — returning an empty result indistinguishable from a genuinely empty one. Unlike a true error, this failure produces no exceptions, no log entries, and no alerts, meaning a user may act on a confidently wrong answer. The researcher proposes a structural fix: scoping the schema to only tables the caller is permitted to read before any SQL is generated, which detected all 10 unanswerable questions in testing at zero additional model cost. This approach reframes the issue as a pipeline design problem rather than a model quality problem, since the silent-denial rate depends on schema design, role permissions, and query mix — not on which AI model is used.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How Keel's Ledger System Prevents Duplicate Job Applications From Ambiguous Submissions

Job application automation tools typically retry failed or timed-out submissions, but unlike fetching a webpage, submitting an application is not idempotent — a retry can create two identical applications for the same role. Keel, a job-application automation system, addresses this by recording every submission intent in a ledger before any network call is made. If a submission ends without a clear server response — due to a timeout, crash, or blank page — the attempt is marked UNKNOWN, a terminal state that is never automatically retried or reclassified. A second attempt for the same role is blocked by the system until a human manually verifies the outcome, such as checking email for an ATS confirmation number. This design trades automation convenience for candidate reputation protection, ensuring no submission is duplicated based on uncertain or missing data.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How Keel's AI Pipeline Handles Unknowns Without Fabricating Answers

Most automated pipelines silently assume success when they encounter missing or unverifiable information, a flaw that can introduce errors or outright fiction into the process. Keel, an AI-driven job application system, treats uncertainty as a distinct signal rather than a formatting issue to be papered over. The platform identifies three specific failure modes: unanswered form questions, unverifiable job posting details, and ambiguous submission outcomes, each handled transparently rather than resolved by assumption. Instead of rejecting or auto-approving uncertain cases, Keel parks them in a review queue with the specific reason attached, allowing the applicant to personally resolve the gap. This design aims to preserve both accuracy and forward momentum, avoiding the twin pitfalls of a system that either lies by filling gaps with guesses or stalls by treating every unknown as a hard rejection.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

ProjectAAL Surpasses 200 Beta Testers in 3 Days, Eyes Gap Between AI Code and Launch

ProjectAAL, a platform designed to help non-technical founders take AI-generated apps to production, grew its beta tester base from 120 to over 200 within three days. The platform targets the gap between generating an app with AI tools and making it genuinely ready for real users, covering areas like backend development, authentication, debugging, and deployment. Alongside the tester growth, the startup secured its first paying customer, completed a pilot client project, and reached $49 in monthly recurring revenue. The team has acknowledged that raw signup numbers matter less than deeper engagement metrics, such as return usage, successful deployments, and willingness to pay. Going forward, ProjectAAL plans to improve backend workflows, integrations, and deployment pipelines, while incorporating expert engineering support directly into the product.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

GitHub HydraFusion Lets Copilot Auto-Select AI Models to Cut Costs

GitHub quietly announced Project HydraFusion in early September, an experimental feature within GitHub Copilot that automatically selects and combines multiple AI models based on the nature of each task. The system uses an orchestration layer offering three routing strategies — Single, Cascade, and Critique — to balance cost and output quality without locking users into one model. In controlled tests against Claude Opus 5, HydraFusion reduced costs by 36–67% across three benchmarks, though quality scores were marginally lower in two of the three tasks. Currently available as a research preview exclusively through Copilot CLI, the feature works on all Copilot plans with no additional fees, charging only standard per-token rates for whichever models are used. GitHub has cautioned that results and behavior may change, and broader support for the main app and VS Code extension is expected to follow soon.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Anthropic's intent.md File Helps Teams Align on Goals Before AI Writes Code

Anthropic's AI-Native SDLC playbook introduces intent.md, a workflow artifact designed to capture the problem and desired outcome of a change before any technical solution is drafted. The file is not a special Claude Code command or auto-discovered configuration, but a human-authored document reviewed by a product owner prior to the design and build stages. It records what needs to change and why, deliberately leaving open questions exposed so the AI agent does not silently make unresolved decisions. A practical example involves a 'Select All' checkbox feature, where the intent file clarifies whether selection should cover the visible page or all filter-matching orders across pages. This approach differs from CLAUDE.md, which provides ongoing project guidance, and spec.md, which defines requirements after intent is agreed upon.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer shares 5 prompt-writing lessons from failed AI video generations

A developer testing AI video generation found that despite a seemingly clear prompt, the camera in a skiing scene barely moved as intended. The experience revealed that vague terms like 'dynamic' or 'cinematic' often fail to convey specific motion instructions to AI models. The developer found better results by describing subject and camera movements as step-by-step paths rather than relying on broad cinematic shorthand. Separating camera movement from subject movement in prompts also proved key to achieving more predictable outputs. These findings led to five practical lessons on writing more precise and effective AI video prompts.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer's AI-Payable API Got 24K Requests in a Week — Only 2 Were Real

A developer launched a paid MCP server offering Korean business verification, priced per call in USDC via the x402 protocol, to test whether AI agents could autonomously discover and pay for an API. Over seven days, the server logged 23,980 requests, but only two were genuine external tool calls, and no third-party payments were received. The bulk of traffic came from crawlers, ecosystem scanners, and liveness bots rather than actual paying agents. The developer identified a core product flaw: the API required a 10-digit Korean business registration number as input, a detail agents typically would not have at the point of needing verification. Key lessons shared include designing around the agent's task rather than raw data access, and ensuring a free tier returns a 402 status so discovery catalogs correctly classify the endpoint as payable.

0 comments Read more at DEV Community
← NewerPage 1004 of 3191Older →