SShortSingh.

Programming

0
ProgrammingDEV Community ·

JQuickCurl Library Simplifies Multipart File Uploads in Java Using curl Syntax

JQuickCurl is a Java library that lets developers send multipart HTTP file uploads using familiar curl -F flag syntax, eliminating the need to manually configure OkHttp or RestTemplate multipart bodies. The library interprets the @ symbol convention to distinguish file paths from plain text form fields, then builds the corresponding multipart request over OkHttp automatically. Developers can upload single or multiple files, mix files with form metadata, and inject dynamic file paths at runtime using variable substitution in annotated interface methods. The approach keeps HTTP implementation details out of service code by expressing upload logic as a standard curl command string in a Java annotation. The library's behavior was demonstrated against the httpbin echo endpoint, which confirmed correct multipart transmission by returning uploaded file data in its response.

0
ProgrammingDEV Community ·

Developer Builds Interactive Lab Site to Host Experiments Beyond a Standard Portfolio

A developer known as joacod has launched 'joacod Lab', a personal web platform designed to host interactive experiments, prototypes, and small tools that don't fit the format of a traditional portfolio or blog. Unlike his static blog, which publishes finished learnings, the Lab is built to support dynamic work such as GPU playgrounds, material studies, and browser-based games with leaderboards. The platform uses a shared shell for navigation and infrastructure while allowing each experiment to maintain its own interface, rendering stack, and interaction model. The developer intentionally avoided over-engineering a plugin framework, following a principle of sharing infrastructure without dictating what each experiment must be. The project evolved from an early dashboard concept into a more flexible architecture that treats each entry as an independent work-in-progress rather than a finished product.

0
ProgrammingDEV Community ·

Plainletter AI Agent Helps Volunteers Decode Official Letters for Non-Native Readers

A developer living in the Netherlands built Plainletter, an AI agent designed to assist library help-desk volunteers in explaining official government and institutional letters to residents who cannot read Dutch. The tool allows a volunteer to photograph or upload a letter, after which it identifies the sender, extracts key details like deadlines, amounts, and reference numbers, and provides a side-by-side explanation in both Dutch and the recipient's language. A built-in verifier cross-checks every extracted fact against the original letter text, refusing to pass any date, amount, or reference it cannot confirm, and sending the rejection back to the model to try again. During testing, the system correctly flagged a misread euro amount differing by just three cents and refused four facts from a deliberately obscured photograph. The project runs on Amazon Bedrock with Strands Agents SDK, costs roughly USD 0.107 per letter reading, and stores only consent-given, masked data for 30 days via AgentCore Memory in Frankfurt.

0
ProgrammingDEV Community ·

Developer finds 5 false claims in own spec; a single tag byte broke a key encoding rule

A software developer audited 18 formal claims in their own project specification using an automated checker that resolves terms against actual definitions. Of the 18 claims, 9 were valid, 5 were invalid, and 4 returned an unknown result. The most concrete failure involved the encoder being labeled a homomorphism, a mathematically strong claim that a single prepended tag byte renders false for every input. Other errors included conflating related but non-interchangeable graph concepts like DAGs, covering relations, and transitive closures, as well as a direct contradiction between lens round-trip laws and a statement 130 lines away in the same document. The developer argues that retaining 'unknown' as a distinct result category is critical, since collapsing it into 'invalid' would misrepresent assumptions as verified facts.

0
ProgrammingDEV Community ·

Multi-Agent AI Setup Matched Single Agent on Every Metric in Controlled Test

A software developer ran a controlled experiment comparing a single LLM-powered support agent against a multi-agent team handling triage, refunds, and knowledge retrieval. Both architectures were tested against identical evaluation scenarios measuring safety, intent accuracy, groundedness, and other properties. The results showed zero difference across all five metrics, with not a single scenario changing its verdict. The multi-agent version added complexity — five types instead of one, 127 lines of code versus 91, and at least twice the model calls per request. The author concluded that splitting tasks across agents changed who invoked the core business logic, but not what that logic does, since the deterministic boundaries remained identical in both designs.

0
ProgrammingDEV Community ·

Bitget DeFi Protocol Rated 7/10 Liquidity Risk on $5.6B TVL: Analysis

A DeFi security assessment dated September 14, 2026, assigned Bitget a liquidity risk score of 7 out of 10, citing significant vulnerabilities across its $5.618 billion in Total Value Locked. The report flagged that 68% of the protocol's TVL is concentrated in a single USDT-USDC liquidity pool on Ethereum L1, creating a major single-point-of-failure. Analysts also identified risks from an admin-controlled upgradeable bridge holding approximately $1.2 billion in escrow, a front-loaded liquidity mining emission schedule, and the absence of an on-chain timelock for protocol upgrades. The protocol's withdrawal system, which batches requests every 30 minutes with no circuit-breaker mechanism, was highlighted as a potential trigger for bank-run dynamics during market stress. While oracle diversification and withdrawal batching were noted as partial mitigations, the overall systemic liquidity risk was rated moderate-high.

0
ProgrammingDEV Community ·

How to Use SpaCy for POS Tagging and Named Entity Recognition in Python

Text tagging is a foundational step in natural language processing that helps identify the role and type of each word in a body of text. Part-of-Speech (POS) tagging classifies words as nouns, verbs, adjectives, and more by iterating through tokens in a SpaCy document. Named Entity Recognition (NER) goes further by identifying real-world entities such as people, places, and organizations using the document's `.ents` property. SpaCy's English model `en_core_web_sm` powers both techniques, relying on machine learning and a transition-based algorithm under the hood. Notably, stripping punctuation or capitalization from input text can reduce NER accuracy, so preserving natural text formatting is recommended for best results.

0
ProgrammingHacker News ·

Cloudflare's Automatic Key Exchange Slashes TLS Retry Requests by 93%

Cloudflare has introduced a feature called Automatic Key Exchange (AKE) aimed at improving TLS handshake efficiency between its edge servers and origin servers. The technology addresses a longstanding inefficiency where mismatched key exchange preferences triggered HelloRetryRequests during TLS connections. Following deployment, the rate of HelloRetryRequests dropped dramatically from 52% to just 3.7%. This improvement reduces latency and computational overhead for origin servers handling HTTPS traffic routed through Cloudflare. The company detailed the technical implementation and results in a post on its official engineering blog.

0
ProgrammingDEV Community ·

Building AI Agents Demands a New Product Mindset Beyond Traditional SDLC

A growing perspective in software development argues that building AI agent-based products requires a fundamentally different approach than traditional software delivery. Rather than designing features users manually interact with, agentic products are built to understand user intent and act on their behalf, with the ability to plan, self-correct, and recover from failures. Architects must now design across three layers — patterns and capabilities, behavior and runtime, and operations — going well beyond simply integrating a large language model. The guiding principle is not maximum autonomy but the right level of autonomy for each specific capability. This shift demands what proponents call an 'Agentic Product Mindset' applied consistently across product, architecture, engineering, and operations phases.

0
ProgrammingHacker News ·

YC-backed Cua seeks Founding Technical GTM Lead

Cua, a startup backed by Y Combinator's P25 batch, is currently hiring for a Founding Technical Go-to-Market Lead role. The position is listed on Y Combinator's official jobs board. This is an early-stage hire, suggesting the company is in its initial growth and commercialization phase. No further details about the role's responsibilities or compensation were publicly disclosed in the listing.

0
ProgrammingDEV Community ·

Milldown app converts document folders to RAG-ready Markdown entirely on your Mac

A developer has released Milldown, a native macOS app that converts PDFs, Word files, spreadsheets, and other document formats into clean Markdown without uploading files to any external server. The app uses Apple's Vision framework for on-device OCR and is powered by Microsoft's open-source MarkItDown engine bundled locally within the app. Key features include folder mirroring, watch folders, RAG chunk export in Markdown or JSONL format, a command-line interface, and token-usage accounting. Benchmarks published by the developer show web pages require 72–95% fewer tokens than raw HTML, while PDFs use 71–84% fewer tokens than native PDF attachments. Milldown requires an Apple Silicon Mac running macOS 14 Sonoma or later, offers a 14-day free trial, and is priced at $39 for a perpetual single-user licence covering up to three Macs.

0
ProgrammingDEV Community ·

Why Argon2id Beats Faster Rivals for Secure Password Hashing

Argon2id is a memory-hard key derivation function designed to be deliberately slow and resource-intensive, making large-scale password-cracking attacks computationally expensive. Unlike older functions such as PBKDF2, which are vulnerable to GPU-based attacks due to their low memory requirements, Argon2id forces each derivation attempt to consume 64MB of RAM, limiting how many guesses an attacker can run in parallel. It is a hybrid of two Argon2 variants — combining GPU-cracking resistance from Argon2d with side-channel attack resistance from Argon2i — without leaving either weakness badly exposed. Configuration parameters like time cost, memory cost, and parallelism must be tuned to balance security against the performance needs of legitimate users. OWASP and other security bodies currently recommend Argon2id as the preferred choice for password storage and key derivation.

0
ProgrammingDEV Community ·

Developer Fixes VS Code Extension That Fired Five API Requests in 15 Seconds

A developer identified and resolved a rate-limiting flaw in their VS Code extension, ds_usage, which displays Claude Code API usage in the status bar. The bug stemmed from each open VS Code window running its own polling timer independently, causing multiple redundant requests — especially when switching windows or clicking the status bar item. The fix introduced a shared cache file via VS Code's global storage, so all windows read from a single source instead of each making separate server calls. A 429 rate-limit response now writes a shared block that all windows respect, and clicks cannot override the cooldown period. The open-source extension, available on the VS Code Marketplace under MIT license, also handles authentication transparently by reading existing Claude Code credentials without storing or logging them.

0
ProgrammingDEV Community ·

How to Accurately Track SaaS Signups in GA4 Without Duplicate or False Events

Developers tracking SaaS signups in Google Analytics 4 should fire the sign_up event only after an account is successfully created, not on button click, to avoid counting failed or duplicate attempts. Google's recommended sign_up event supports a method parameter — such as 'Email' or 'Google' — to describe how the account was created. A single source of truth should own the event to prevent double-counting from components like both a signup form and a welcome page triggering it simultaneously. Separate events should be used for distinct milestones: the button click, account creation, and the user's first meaningful product action each answer different analytical questions. Before implementation, teams should define success criteria clearly, validate tracking using GA4's Realtime or DebugView tools, and ensure no personally identifiable information is included in event parameters.

0
ProgrammingHacker News ·

Amazon Science Explores Why ML Research Agents Resist Overfitting

A new article published on Amazon Science examines a counterintuitive phenomenon in machine learning research agents: their apparent resistance to overfitting. Overfitting is a common problem in ML where models perform well on training data but poorly on unseen data. The piece investigates why autonomous research agents, despite repeated exposure to similar problem structures, do not seem to fall into this trap. The discussion has attracted attention in the machine learning community, sparking early debate on Hacker News. The findings could have broader implications for how researchers design and evaluate AI-driven scientific discovery systems.

0
ProgrammingDEV Community ·

How to Build an AI Customer Support Chatbot That Also Upsells Products

Developers can build a fully functional AI customer support chatbot in roughly 4 to 6 hours using OpenAI's GPT-4o, the n8n workflow engine, and a Pinecone vector database. The bot handles common support queries and can suggest relevant products or upgrades to customers without any human agent involvement. It uses retrieval-augmented generation (RAG) to pull product data from a vector store, enabling contextually accurate responses. Customer interactions can be delivered via Twilio SMS, WhatsApp, or a Bubble-powered web widget. The estimated cost is largely pay-as-you-go, with free tiers available for most tools including Pinecone and n8n self-hosted.

← NewerPage 1255 of 3229Older →