SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World

Programming

LatestTopDiscussed
0
ProgrammingDEV Community ·

Developer Builds Exoroute, a Rust-Based AI Router Using Just 80 MB RAM

A developer created Exoroute, a lightweight AI router built with Rust, after finding that its predecessor Omniroute consumed over 2.8 GB of RAM for just 11 connections. Exoroute addresses these resource efficiency concerns by running 14 concurrent users at roughly 1.3% CPU and 80 MB of RAM. The developer notes these benchmarks were recorded on a personal Linux device and may vary across different systems. Released as version 0.1.0, Exoroute is available for installation via npm and can be explored at exoroute.dev. The developer is actively seeking community feedback on this initial release.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

14 AI Agents Screened 120,000 Influencers to Pick 30, Boosting CTR by 2.7%

A multi-agent AI system comprising 14 specialized agents was deployed in a real marketing campaign to automate influencer selection and ad execution end-to-end. The system scanned approximately 120,000 influencer profiles and narrowed them down to 30 using multi-dimensional scoring that factored in content style, audience alignment, conversion history, and pricing efficiency. Each selected influencer received a customized 60-point strategy brief tailored to their format and audience, rather than a generic campaign document. The automated pipeline replaced a traditionally fragmented manual workflow involving multiple platforms, spreadsheets, and personnel, delivering a reported 30% efficiency gain. The campaign ultimately achieved a 2.7% improvement in click-through rate, with no manual execution steps required between influencer discovery and ad settlement.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer builds keyless USDC deposit address, nearly derailed by 9,999-block RPC limit

A developer built a keyless crypto deposit system called Pigeonhole on Arc, Circle's L1 blockchain where USDC serves as the native currency for gas fees. Instead of generating private keys, the system uses Ethereum's CREATE2 opcode to deterministically derive deposit addresses, with funds swept to a treasury via a self-destructing contract in a single transaction. The project's entire payment-status logic relies on querying blockchain event logs, which exposed a critical flaw discovered the night before submission. Testing revealed that Arc's public RPC rejects log queries spanning 10,000 or more blocks, capping requests at 9,999 blocks — a limit that would have frozen invoice statuses within hours of deployment. The bug was caught during a pre-submission audit and subsequently fixed before the live version went public.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Why Additive Risk Scoring Fails Against High-Confidence Threat Intel Signals

A cybersecurity team discovered that their purely additive risk-scoring model was dangerously under-reacting to high-confidence threat intelligence hits. In one case, an IP confirmed by ThreatFox as active Cobalt Strike command-and-control infrastructure scored only 33 out of 100, landing in the 'LOW' severity tier and bypassing automated review. The root problem was that the additive model treated a definitive malicious confirmation as just another point contributor, equal in weight to weak probabilistic signals like unusual user-agent strings. The team resolved this by introducing a score floor rule: any single authoritative confirmation of malicious activity — from sources like ThreatFox, OTX, GreyNoise, or AbuseIPDB — automatically sets the event score to a minimum of 75, triggering full analysis. The broader takeaway is that additive scoring suits weak-to-moderate signals but must be paired with floor rules when any signal is meant to convey near-certainty of a threat.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Why verifying AI outputs is harder and more critical than generating them

A developer who has built multiple AI infrastructure tools argues that trustworthy AI systems depend not on model confidence but on enforceable, auditable constraints. The core principle is that models should be structurally prevented from accessing data they shouldn't see, rather than simply instructed to behave correctly via prompts. Tools like vaultrag and Bridgekit implement permission-aware retrieval and narrowly scoped agent access to enforce these boundaries at the system level. Beyond access control, the developer stresses the need for tamper-evident logs and verifiable records — such as signed retrieval receipts — so that what a model actually did can be reconstructed from evidence rather than assumed. Irreversible actions, including sending messages or deleting records, are flagged as requiring a mandatory human approval step before execution.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Telegram's username field can be empty; here's how to extract the correct handle

Telegram's API can return an empty string for the username field of a chat, even for high-profile accounts like founder Pavel Durov's, because accounts with multiple public handles move them into a separate usernames array. Each entry in that array carries active and editable boolean flags, and only the entry marked both active and editable represents the canonical, owner-controlled handle. Developers who build links using the plain username field or a numeric chat ID risk generating URLs that appear valid but resolve to nothing for users. A three-step fallback function — checking for an active-and-editable handle first, then any active handle, then the plain field — correctly covers both single-handle and multi-handle API response shapes. The issue commonly surfaces when a channel renames itself and retains old aliases, silently emptying the username field without triggering obvious errors.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

1 in 88 WordPress Sites Near French Town Found Secretly Serving SEO Spam

A web agency based in Vienne, France, scanned 1,084 WordPress-powered business websites within a 25 km radius of the town to assess how many were silently compromised. Using two external detection methods — cloaking analysis and sitemap inspection — researchers identified 12 infected sites, a rate of 1.13% of the 1,066 fully analysed. The cloaking test compared pages served to regular browsers versus Googlebot, while the sitemap check scanned URLs for spam vocabulary. Notably, the two methods caught entirely different sets of six sites each, meaning either check alone would miss half the infections. Researchers noted the true infection rate is likely higher, as some sites blocked the test crawler and compromises recur over time.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Tests Jev Probability Tool Against LLMs for Classifying Cinema Listings

A developer behind Clusterflick, a London cinema listings aggregator, has been using a language model for over a year to categorise 450–600 daily listings into one of ten types, such as movie, talk, or event. The task is complex because many listings, like a Buster Keaton society screening at the Cinema Museum in Kennington, do not fit neatly into a single category. The developer recently trialled Jev, a non-generative tool that returns probability distributions over structured options rather than producing text. Unlike the LLM approach, Jev evaluates multiple yes/no questions in parallel within a single request, allowing underlying classification factors to be tested directly. The experiment explored whether Jev could match or improve on a year's worth of prompt refinement, given that the text explanations produced by the LLM were never actually used.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Shares Weekly Learnings Across ML, Backend, and Web Security Basics

A developer recently documented self-directed learning across three technical domains: machine learning, backend development, and web application security. On the ML front, they studied supervised, unsupervised, and reinforcement learning, along with exploratory data analysis and linear regression. Their backend exploration covered foundational concepts such as domains, subdomains, and the HTTP protocol. They also examined Cross-Site Scripting (XSS) attack types — stored, reflected, and DOM-based — with a focus on how stored XSS exploits vulnerabilities and how developers can defend against them.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

ShortGeek Converts Articles into Narrated Vertical Videos Entirely Offline

A developer has released ShortGeek, a locally run tool that transforms written articles or RSS feeds into narrated vertical short-form videos without any cloud uploads or subscriptions. The tool handles the full pipeline: drafting a script, generating narration, timing captions at the word level, and rendering the final video using Python, FastAPI, ffmpeg, and local AI models. Word-level caption timing was the most technically challenging part, ensuring captions sync accurately with the narration rather than displaying full sentences at once. A built-in render queue allows users to batch-process an entire RSS feed overnight, since individual renders take considerable time. The developer notes that ShortGeek acts as a drafting assistant rather than a finished production tool, and the installer is not yet code-signed, which may trigger a SmartScreen warning on first run.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

Developer builds AI-assisted Bitcoin full node in Rust using libbitcoinkernel

A developer has shared bitcoin-rs, an independent Bitcoin full node being built in the Rust programming language with heavy use of AI tools for implementation. The project was posted to Hacker News as a community showcase, garnering early attention. The developer argues that Bitcoin is a strong candidate for AI-assisted development because implementations can be cross-checked against well-established external references. To ensure compatibility and correctness, bitcoin-rs uses libbitcoinkernel as a primary verification oracle. The project is currently in active development and open for community feedback.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

Video Shows AI Agent Running 50 Social Media Accounts Simultaneously From One Workstation

A video posted on X on September 19, 2026, appears to show a single workstation operating fifty social media accounts simultaneously using a Chinese AI agent, with no human involvement. The footage depicts dozens of emulated smartphone screens generating posts, replies, and likes around the clock, illustrating what researchers call 'dead internet theory' in practice. First proposed in a 2021 forum essay, the theory argued that most online activity was quietly becoming non-human, a prediction that has since gained empirical support. Cloudflare reported in June 2026 that automated traffic had surpassed human traffic, accounting for roughly 57.4% of web page requests — more than a year ahead of its CEO's own forecast. HUMAN Security's 2026 benchmark further found that action-taking AI agent traffic grew approximately 7,851% year over year, signaling a fundamental shift in who — or what — now drives activity online.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

ALIS 2.0.0 Lets Players Fly Over Real-World Kazan and Manhattan in UE5

The open-source game ALIS has released version 2.0.0, introducing two real-world cities — Kazan and Manhattan — selectable directly from the main menu. Players can fly over large city territories featuring terrain, water, roads, greenery, and buildings reconstructed from actual geographic data. The release is built on Unreal Engine 5.8 and uses a deterministic pipeline that generates both cities from real-world geographic sources. Developers gain access to automated systems covering world generation, streaming, gameplay routes, and performance testing. The project remains an early prototype, and the download, source code, and signed checksums are available on GitHub.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Profullstack publishes machine-readable tech stack guide to standardize AI agent coding

Profullstack has published a structured technology stack reference at profullstack.com/stack, available in Markdown and JSON formats so AI agents can read it before generating code. The company manages around 171 active repositories with human commits since June 2025, and different project types — web apps, APIs, CLIs, desktop, and mobile — each follow a distinct set of technologies. The stack page organizes these as named combos, mapping each use case to its recommended tools, layers, and live example repositories. It also lists universal defaults covering deployment, authentication, payments, and testing, along with 12 house rules and a section on tools explicitly excluded from the stack. The content was derived from a survey of 251 repositories and corrects inaccuracies found in an older internal skills document.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer builds unofficial browser to preview and pick Impeccable design directions

A developer built Impeccable Worlds, an unofficial companion tool for the AI design system Impeccable, and published it over a weekend as an open-source project. The browser-based catalog lets users search, filter, and preview human-reviewed design directions called 'worlds' before committing one to an AI coding agent. The tool addresses a gap in the existing Impeccable workflow, where design directions are typically selected through an automated roll rather than deliberate visual comparison. Users can filter worlds by type, save favorites, and copy a direction prompt to lock in a choice and prevent the agent from re-rolling it. The project is self-described as incomplete, unofficial, and unaffiliated with Impeccable's creator Paul Bakaus, built with React 19, TypeScript, and Vite.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

Opinion: Why Writers Should Avoid Delegating Their Work to AI

A blog post by Paul Bakker argues against using artificial intelligence tools to produce written content. The author contends that writing is a cognitive process that builds thinking and clarity, not merely a means to produce output. Delegating writing to AI, the piece suggests, robs the author of the intellectual benefits that come from the act of composing. The post sparked discussion on Hacker News, earning 23 points and 8 comments from readers engaging with the argument.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

Faux Pas Atlas lets locals collectively judge travel etiquette, skipping verdicts

Faux Pas Atlas is a crowd-sourced travel etiquette platform built for the Sanity Challenge that answers whether a behaviour is considered rude in a given country or context. Unlike traditional travel guides, the site stores no verdict in its database — consensus is computed live from weighted user attestations, with locals counting double and visitors half. A claim is only considered settled when at least five people have responded, three of them local, and 70% of weighted votes agree; when locals disagree, the site displays the split rather than choosing a side. Users can propose new claims, dispute settled ones, and take country-specific quizzes built entirely from locally verified answers. The open-source project uses Sanity Studio, Astro, and Cloudflare Workers, with a moderator dispute board integrated directly into the Sanity Dashboard.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Captions Bro app combines video editing, auto-captions and teleprompter in one tool

Developer Andrew has launched Captions Bro, a video editing app for iPhone and iPad, after six months of development. The app targets short-form content creators on platforms like TikTok, Reels, and YouTube Shorts by consolidating recording, captioning, and editing into a single workflow. Key features include on-device speech transcription, a built-in teleprompter, multi-track video editing, keyframe animation, and support for vertical video formats. Unlike cloud-based tools, all transcription is processed locally on the device, removing the need for an internet connection. The app is now available on the App Store, and the team is seeking feedback from creators to guide future development.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Stirling-PDF Lets You Run 50+ PDF Tools Privately on Your Own Server

Stirling-PDF is a self-hosted, open-source PDF toolbox that processes documents entirely on your own server, ensuring files never reach third-party services. Version 2.14.2 can be deployed via Docker Compose behind a Traefik reverse proxy, exposing over 50 tools including merging, compression, format conversion, OCR, redaction, and watermarking. The application is stateless by design, meaning it does not permanently store uploaded files, with only configuration and OCR language data saved in mounted volumes. Access is protected by a login system that forces a password change on first use, replacing the publicly known default credentials. The tool integrates well with self-hosted document platforms such as Paperless and Nextcloud.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Builds Data Analyst AI Agent Using Google ADK at Kisumu Event

At a recent Build with Google AI event in Kisumu, a developer built a Data Analyst Agent using Google's open-source Agent Development Kit (ADK), a Python framework for creating autonomous AI agents. The project demonstrated agentic workflows, where complex tasks are split across specialized units coordinated by an orchestrator, rather than relying on a single AI prompt. During local testing, the developer encountered two configuration errors — an invalid model endpoint and a formatting typo in the model name string — both of which were resolved by correcting the configuration files and refreshing authentication. The Model Context Protocol (MCP) was highlighted as a key standard that allows agents to connect uniformly to external databases and tools without custom integrations. Key takeaways from the build include the importance of exact model name syntax, using MCP for tool standardization, and leveraging local tracing tools like ADK's built-in web UI to debug API issues efficiently.

0 comments Read more at DEV Community
← NewerPage 1053 of 3197Older →