SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World

Programming

LatestTopDiscussed
0
ProgrammingDEV Community ·

Network Engineer Joins DEV to Document Hardware, Software, and Infrastructure Projects

A network engineer going by the handle 1ndiehacker has joined the DEV Community platform to share their work on personal tech projects. Their interests span networking, embedded systems, wireless communication, microcontrollers, and security. One ongoing project evolved from an attempt to route IP traffic over a low-bandwidth wireless link, which failed but sparked a new idea focused on transmitting only the specific data needed. The engineer plans to document not just completed projects but also the failures, debugging sessions, and design decisions behind them. They are looking to connect with others interested in networking, embedded systems, and hands-on building.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Builds Accessible Browser Scaling Game Using Astro and React Islands

A developer has created 'Size It Up Party', a browser-based game where players visually estimate size ratios between silhouettes rather than typing numeric answers. The project uses Astro for static pages and a focused React island solely for the interactive gameplay area, keeping the site fast, crawlable, and accessible. Scoring, validation, and normalization logic are housed in a separate TypeScript package, making the rules reusable across Daily and Unlimited game modes and testable independently of the DOM. Accessibility was a core design priority, with keyboard navigation, labeled buttons, and touch-friendly controls ensuring all users can interact without relying on precise pointer movement or color alone. The build demonstrates that a single interactive component can coexist with ordinary static content without requiring a fully client-rendered application.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

AI Made the Same 3 Planning Mistakes Across 170 Goals, Regardless of Model Size

A developer built a small engine called PlannerCritic that paired one LLM to write plans with a second to review them, then tested it across 170 change-planning goals spanning 40 domains at a total cost of $0.49. Out of 132 concrete blockers identified, 121 fell into just three recurring failure categories: unverified dependencies, unsafe sequencing, and weak rollback logic. Switching to a more powerful model like GPT-4o produced better-written plans but did not eliminate these structural defects. The author found that deterministic validation gates — which ran without any LLM calls — were more effective at catching bad plans than upgrading the underlying model. The key takeaway is that AI planning failures in multi-step tasks are structural problems, not parameter or model-size problems.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Real Bot lets Mac users run persistent local AI teammates with multi-bot support

A developer has released Real Bot, an open-source macOS application that enables users to run multiple AI bots locally on their machines. The tool supports named bots with assigned duties, group chats, @mentions, and task handoffs between bots within a shared workspace folder. It is compatible with any OpenAI-compatible endpoint, allowing users to bring their own API keys, which are stored securely in macOS Keychain. A built-in approval system pauses risky actions until the user confirms them, and MCP support is available via stdio or Streamable HTTP. The project is MIT-licensed and currently in alpha, available as an unsigned .dmg file, with no affiliation to xAI or Grok.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Critical Citrix NetScaler Flaw CVE-2026-19490 Allows Authentication Bypass via SAML

A critical vulnerability, CVE-2026-19490, has been identified in Citrix NetScaler ADC and NetScaler Gateway, carrying a CVSS v4.0 score of 9.3. The flaw stems from a logic error in the SAML HTTP-Redirect binding handler that fails to enforce signature verification, allowing attackers to submit unsigned assertions and obtain valid sessions without any credentials or cryptographic material. Affected versions include NetScaler ADC and Gateway 14.1 before 14.1-73.32 and 13.1 before 13.1-63.21, along with corresponding FIPS and NDcPP builds, while end-of-life branches 12.1 and 13.0 will not receive patches. Citrix published advisory CTX696939 on August 19, 2026, and CISA added the vulnerability to its Known Exploited Vulnerabilities catalog on September 9, 2026. Because NetScaler operates as boundary infrastructure, a successful exploit could expose all resources behind the gateway, and the attack leaves no signature-failure log entries, making detection particularly difficult.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How to Add Custom Text to PDF Files Using Python's Free Spire.PDF Library

Developers can add custom text to existing PDF documents in Python using the Free Spire.PDF library, which requires no local Adobe components. The library supports a range of styling options including custom fonts, RGB colors, transparency, and text rotation, making it suitable for use cases like watermarks, annotations, and page numbering. Installation is straightforward via pip and is compatible with Python 3.7 and above across Windows, macOS, and Linux. The basic workflow involves loading a PDF, selecting a target page, configuring font and brush settings, and drawing text at specified coordinates before saving the file. Advanced customization is also available through multiple font initialization methods, including CJK fonts, system TrueType fonts, and built-in PDF standard fonts.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

33,000 LiteLLM Gateways Exposed Online After Auth Bypass Flaw CVE-2026-59822

A critical authentication bypass vulnerability, CVE-2026-59822, was discovered in LiteLLM's MCP Streamable HTTP endpoint and patched in version 1.84.0, with CISA adding it to its Known Exploited Vulnerabilities catalog on September 2, 2026. A ZoomEye scan conducted on September 16, 2026 identified 33,005 internet-facing instances fingerprinted as LiteLLM, highlighting the broad exposure of this AI gateway product. The flaw allowed unauthenticated callers to bypass key checks via a fail-open fallback that substituted an empty auth object instead of rejecting invalid requests. Security researchers at Wiz and Microsoft reported in September 2026 that attackers, linked to the Qilin ransomware group, were chaining this vulnerability with other flaws to achieve code execution on AI infrastructure. Because LiteLLM acts as a centralized proxy holding upstream provider API keys and tool access, a single compromised gateway can expose significant credential material and internal resources.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Factory Method Pattern in Laravel: Stop Scattering 'new' Across Your Codebase

The Factory Method design pattern addresses a common problem in Laravel applications where direct object instantiation using 'new' is scattered across multiple files and controllers. This tight coupling makes codebases fragile, as changes to a class constructor — such as adding a logger or cache manager — require updates in dozens of places. The pattern centralizes object creation into dedicated factory classes or methods, so client code depends on abstractions rather than concrete implementations. This approach improves testability by allowing mock factories to be injected, and supports the Open/Closed Principle by letting developers add new types without modifying existing code. Using a coffee shop analogy, the article illustrates how a centralized 'machine' handling creation logic is far more maintainable than training every barista individually for each recipe change.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Floci emerges as free open-source AWS emulator after LocalStack ends community edition

Floci is a free, MIT-licensed local AWS emulator that runs services like S3, Lambda, DynamoDB, and SQS inside Docker, requiring no account or authentication token. The tool gained relevance after LocalStack retired its open-source Community edition in March 2026, when version 2026.03.0 began requiring a paid account token to run. Built on Quarkus and compiled as a GraalVM native binary, Floci claims a roughly 24-millisecond startup time and 13 MiB idle memory footprint, significantly lighter than LocalStack's reported figures. Lambda functions execute inside real Docker containers using genuine AWS runtime images, and services such as API Gateway, Cognito, and IAM-authenticated databases are included without paywalls. Teams using LocalStack in local development or CI pipelines can migrate by replacing a single image reference in their existing Docker Compose configuration.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

ASCII City: A Browser-Based City Simulation Rendered in Text Characters

ASCII City is a web-based project accessible at asciicity.live that renders a city simulation using ASCII text characters. The project was shared on Hacker News, where it received 10 upvotes at the time of posting. The submission attracted no comments, suggesting it is a relatively new or niche submission. The project appears to be an interactive or visual experiment combining classic ASCII art aesthetics with city-building or simulation concepts. No further technical details about the creator or development background were provided in the original post.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

How to Choose the Right Multimodal AI API for Your Workload in 2026

Selecting a multimodal AI API in 2026 requires matching the platform to the specific workload rather than simply comparing model counts or opting for a single provider. Key questions include what output types are needed, whether models are commercial or custom-deployed, and how billing units like tokens, megapixels, or generated seconds align with actual usage. Four platform categories stand out: cross-provider gateways for mixed workloads, Replicate for open-model experimentation, fal.ai for high-volume media generation, and Google Vertex AI for teams already operating within Google Cloud. Vertex AI is particularly suited to regulated or governance-heavy environments, where Google's own models — Gemini, Imagen, Veo, and Lyria — integrate with existing IAM policies and regional controls. Pricing figures cited, such as Gemini 3.8 Flash at $0.75 per million input tokens, reflect a September 2026 snapshot and should be independently verified before being used in cost models.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

IBM's Chunkless RAG Challenged: Does Ditching Document Chunks Actually Help?

IBM has been promoting a 'Chunkless RAG' approach that uses AI agents to navigate document structure — preserving headings, tables, and hierarchies — instead of splitting text into fixed-size chunks for embedding. Critics argue the method is overhyped, noting that real-world document corpora — including scanned PDFs, legal contracts, and Slack exports — are too messy for reliable structure parsing. When a parser fails to extract a clean document tree, the navigating agent simply traverses noisy data with added latency, swapping one failure mode for another that is harder to detect. Analysts also point out that the core retrieval problem in most pipelines is poor query-to-passage semantic overlap, which Chunkless RAG does not inherently solve. Techniques like hypothetical document embeddings, query expansion, and hybrid BM25-plus-dense-vector retrieval are seen as more reliably moving the needle on retrieval precision.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer builds TGDown Chrome extension for one-click Telegram Web media downloads

A developer has released TGDown, a Manifest V3 Chrome extension built with TypeScript and Vue that adds download controls for images and videos directly within Telegram Web. The extension supports both Telegram Web interfaces and offers one-click downloads, batch downloads from search results, and a popup panel listing detected media. TGDown operates across three execution contexts — a content script, a page-context script, and extension pages — to stay isolated from Telegram's internal code while using its existing authenticated session. During development, a critical dark-mode bug caused the entire Telegram page to flash white due to a full-screen iframe rendering the browser's default white canvas before initializing. The developer resolved the issue by replacing the iframe approach with a Shadow DOM host, which maintains CSS isolation without introducing a separate document rendering surface.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Why Video Background Removal Fails and How Temporal AI Models Fix It

Developers commonly misapply image-based background removal tools to video, treating each frame as an independent photo rather than part of a continuous temporal sequence. This approach causes visible flickering, ghosting on fast movements, and inconsistent edge detection between frames. Architectures like Robust Video Matting (RVM) address this by using recurrent neural networks that carry memory across frames, enabling stable real-time matting at resolutions up to 4K. A separate but related pitfall is exporting transparent video as H.264 MP4, which does not support an alpha channel, causing transparent pixels to render as black; formats like WebM or ProRes 4444 MOV are needed for true transparency. Understanding both the temporal processing model and format constraints is essential for developers building reliable video segmentation pipelines.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

How to Self-Host a Dockerized App on WSL2 Using a Cloudflare Tunnel

A developer built a proof-of-concept to-do app using Node.js, Express, and Postgres running inside Docker on WSL2, made publicly accessible via a named Cloudflare tunnel on a custom domain. The app uses a per-browser cookie hash instead of a login system, and automatically clears its database every two hours. Two deployment approaches were tested: a bare Dockerfile connecting to a Windows-hosted Postgres instance, and a docker-compose setup with its own Postgres and a crond sidecar for scheduled cleanup. The project revealed that using raw Docker Engine inside WSL2, rather than Docker Desktop, broke host networking assumptions and required rethinking the architecture. The write-up shares practical lessons on WSL2 Docker setup, pg_cron limitations on Windows, and idiomatic container patterns for periodic tasks.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Critical NetScaler SAML Bypass CVE-2026-19490 Scores 9.3, Exploits Already Observed

A critical vulnerability tracked as CVE-2026-19490 affects Citrix NetScaler ADC and NetScaler Gateway, allowing attackers to forge SAML assertions and obtain valid sessions without a legitimate signature. Citrix published advisory CTX696939 on August 19, 2026, and CISA added the flaw to its Known Exploited Vulnerabilities catalog on September 9, 2026, confirming active exploitation in the wild. The vulnerability carries a CVSS v4.0 score of 9.3 and is particularly dangerous because compromised sessions appear legitimate to downstream systems, leaving no obvious trace in authentication logs. Affected builds include NetScaler ADC and Gateway versions 14.1 before 14.1-73.32 and 13.1 before 13.1-63.21, with no workaround available — patching to a fixed build is the only complete remedy. End-of-life branches 12.1 and 13.0 will not receive fixes and should be replaced or isolated, while a ZoomEye scan identified over 239,000 publicly reachable NetScaler instances, underscoring the scale of potential exposure.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

ExitOnForwardFailure=yes confirms SSH tunnel setup, not database availability

A technical analysis published on DEV Community clarifies a common misconception about SSH tunneling: the OpenSSH option ExitOnForwardFailure=yes only confirms that a local forwarding listener was successfully established, not that the service behind the tunnel is reachable. Testing conducted on macOS with OpenSSH 10.3p1 and Node.js 25.6.1 demonstrated that an SSH process can continue running even when the destination port returns a connection refused error. The experiment showed that a local TCP connection to the forwarding port succeeds independently of whether the target service is actually up. Developers troubleshooting PostgreSQL connection failures through SSH tunnels are advised to diagnose each layer separately: the local listener, the SSH transport, the destination connection, and the database protocol itself. A successful local TCP probe alone cannot verify a query, database identity, or authentication credentials, so the final check must use a proper database client with correct credentials and TLS settings.

0 comments Read more at DEV Community
0
ProgrammingHacker News ·

Jev Ultrafast: Open-Source Browser Agent With Dynamic Indexed Action Space

Jev Ultrafast is a browser automation agent built on the browser-use framework, recently shared on Hacker News. It features a dynamic, indexed action space designed to make browser interactions faster and more efficient. The project is open source and hosted on GitHub under the browser-use organization. The submission received 10 points on Hacker News, though no public comments had been posted at the time of sharing. The tool appears aimed at developers looking to build or experiment with ultrafast browser agents.

0 comments Read more at Hacker News
0
ProgrammingDEV Community ·

Two AI agents fixed 12 code issues overnight using 158 million tokens at steep cost

On September 15, 2026, a team at ShortSingh-style dev shop deployed one AI coding agent to resolve twelve design flaws in a web service, completing the work overnight without any human typing code. A second AI agent from a different vendor independently rebuilt and verified the output the following morning, confirming 100 of 100 tests and 73 of 73 browser checks passed with a byte-identical release package. The entire run consumed 158 million tokens over roughly 13 hours, yet 95.4% of those tokens were re-reads of already-cached context, with models writing only 0.42% of new text. To reduce costs, the team routed tasks by complexity — using cheaper models for routine work and premium models only for demanding reasoning — cutting token spend by 54% compared to running everything on the top-tier model. Despite the efficiency gains, the team acknowledges the overall cost remains unacceptable and is actively working on further optimizations, including zero-token telemetry and stricter budget controls.

0 comments Read more at DEV Community
← NewerPage 1186 of 3224Older →