SShortSingh.

Programming

0
ProgrammingDEV Community ·

Dev Team Learns Hard Lessons After Building Multi-Repo Project with Git Submodules

A development team designed their project around five separate Git repositories and submodules from day one, prioritizing CI isolation, access control, and clean history. In practice, the setup demanded strict two-step push discipline — once inside the submodule and once in the parent repo — and forgetting either step caused silent failures that only surfaced during fresh CI checkouts. On one occasion, three submodules simultaneously had unpushed local commits, while a fourth required an interactive rebase to recover. The core 'engine' repository also caused CI conflicts by behaving differently as a standalone repo versus as a nested submodule, breaking path-resolution logic and missing cross-repo dependencies. These recurring failures pushed the team to move beyond one-off patches and establish a shared convention for managing the dual-context build environment.

0
ProgrammingDEV Community ·

Python Data Types Explained: Practical Guide to int, list, dict, and More

Python uses distinct data types — including integers, floats, strings, lists, tuples, dictionaries, sets, booleans, and None — to categorize and manage different kinds of values in a program. Each type serves a specific purpose: lists are mutable collections, while tuples are immutable; dictionaries store key-value pairs useful for structured data like JSON; and sets automatically eliminate duplicate values. Floats can produce minor precision issues due to internal binary representation, which matters in calculations requiring exactness. Booleans represent True or False and are widely used in conditions and data filtering, while None signals the absence of a value. Understanding these data types is foundational for working with functions, APIs, databases, and data analysis in Python.

0
ProgrammingDEV Community ·

Why Viewport Visibility Does Not Equal a Valid Survey Impression

A developer behind RetroPrompt has outlined the technical and methodological care required when measuring survey exposure on a webpage. The article explains that a page view alone cannot confirm a visitor actually saw an optional question, since the question appears below the main tool and can be missed entirely. Using the browser's IntersectionObserver API, exposure is defined as at least 50% of the survey panel entering the viewport, though this still does not guarantee the visitor read or engaged with it. The piece warns against conflating unanswered sessions with a preference for the default option, and stresses that first-choice selections and silence must be reported separately. It also cautions that self-selected survey responses do not indicate demand, and that small sample sizes should always be accompanied by an acknowledgment of uncertainty.

0
ProgrammingDEV Community ·

AI Boosts Coding Productivity but May Shortcut the Learning Process

A developer perspective published on DEV Community argues that while AI tools have significantly accelerated coding workflows, they risk allowing developers to become productive before they become genuinely skilled. The author draws a key distinction between asking AI to explain an error versus simply asking it to fix one, noting that the latter can eliminate the struggle that builds deeper understanding. Using a hypothetical comparison, the piece illustrates how a developer who spends hours working through a problem may retain more durable knowledge than one who resolves it in minutes via AI. The author is not calling for abandoning AI tools, but rather urging developers to distinguish between unnecessary struggle worth skipping and productive struggle worth preserving. The core concern is that modern development increasingly rewards task completion, and AI excels at completion — potentially masking gaps in foundational understanding.

0
ProgrammingDEV Community ·

Developer builds free tool to increase image file size for Indian exam form uploads

A developer has launched IncreasePhotoSize, a browser-based tool designed to increase image file sizes in KB — the opposite of conventional compression tools. The tool addresses a common problem faced by millions of Indians filling online forms for bank exams, railway recruitment, and college admissions, where uploaded photos are rejected for being below a minimum file size threshold. Unlike most image tools, it distinguishes between file size, pixel dimensions, and DPI, adjusting JPEG quality and dimensions until the output meets a target range. All processing runs client-side using the browser's Canvas API, meaning no images are uploaded to any server — a deliberate privacy choice given users often submit ID photos and signatures. The tool is free, requires no sign-up, adds no watermark, and includes presets tailored to specific Indian government exam portals.

0
ProgrammingDEV Community ·

Developers Build Graph-Based Agentic Fraud Investigation Tool Using TigerGraph

A team participating in the TigerGraph × HHGoa 2026 hackathon challenge built an agentic fraud investigation system using TigerGraph Savanna as the graph layer. The solution organizes transaction, customer, and card entities through interconnected relationships to surface fraud-related evidence. A custom Python agent applies rule-based fraud pattern detection and heuristic risk scoring to generate structured investigation decisions for each case. The system successfully processed all 20 benchmark cases, from HHG-001 to HHG-020, with results stored as individual JSON files. Future development plans include full TigerGraph MCP integration, LLM-assisted reasoning, and more advanced fraud detection models.

0
ProgrammingDEV Community ·

Key Security Controls Every Custom AI Application Needs in Production

As organizations deploy custom AI applications built on large language models and retrieval-augmented generation, security experts warn that making an app functional is very different from making it secure. Every incoming request must be treated as untrusted input, with systems in place to detect prompt injection, jailbreaks, and policy-bypass attempts before they escalate. Output filtering is equally critical, since even legitimate queries can trigger responses that inadvertently expose sensitive data, credentials, or internal business information. Security layers should be deployable via SDKs, middleware, or API gateways so teams can protect existing workflows without rebuilding their entire architecture. Centralized monitoring and policy-driven responses — such as redacting, blocking, or alerting — are recommended to shift AI security from one-time controls to an ongoing capability.

0
ProgrammingDEV Community ·

OpenClaude Offers Free, Multi-Provider Alternative to Anthropic's Claude Code Terminal Agent

OpenClaude is a free, open-source terminal coding agent forked and substantially reworked from Anthropic's Claude Code codebase, published on GitHub since April 2026 and now carrying over 33,000 stars. Unlike Claude Code, which is Anthropic's first-party product bundled into paid plans starting at $17 per month, OpenClaude costs nothing to install and supports 24 provider routes including local models via Ollama with no API key required. Users pay only their chosen model provider's token rates, making it a practical option for those without easy access to international payment methods or subscription plans. Claude Code remains the stronger fit for existing Claude Pro or Max subscribers, as terminal usage draws from the same shared usage pool already covered by their plan. The choice between the two tools ultimately comes down to cost structure and model flexibility rather than any proven difference in code quality, as no public benchmark currently compares the two agents directly.

0
ProgrammingDEV Community ·

LinkedIn CEO: AI Engineering Jobs in India Growing 51% Year-on-Year

LinkedIn CEO Dan Shapero, during his first India visit in the role, said AI engineering jobs on the platform are growing 51% annually, citing LinkedIn's own hiring data. He highlighted healthcare, education, startups, robotics, and data-centre operations as the sectors driving demand, with data-centre hiring having tripled over the past decade. India's share of overseas R&D hiring by US companies rose sharply from 27% in 2019 to 41% in 2025, underscoring the country's growing importance as a global talent hub. Shapero also noted that skills listed in Indian job postings change at roughly 12% per year, signalling rapid shifts in what employers require. He attributed the global hiring slowdown more to elevated interest rates than to AI-driven job displacement.

0
ProgrammingDEV Community ·

Fable 5.1 Tops Real-SWE Benchmark at 38.8%, but Costs Nearly $7 Per Task

Specific Labs released the Real-SWE benchmark in September 2026, evaluating eight AI model-and-harness combinations across ten tasks drawn from real enterprise codebases covering billing, tax, migration, and infrastructure work. Fable 5.1 paired with Claude Code achieved the highest resolution rate at 38.8% and an estimated cost of $6.96 per rollout, largely due to its high token output averaging over 80,000 tokens on several tasks. GPT-6 Astra via Codex CLI ranked second at 33.8% for $4.67, while Gemini 3.8 Flash via Gemini CLI offered the best value at 31.2% for $2.50 per rollout. Confidence intervals among the top three overlap significantly, meaning the performance differences are directional rather than definitive. Notably, longer agent runtimes did not improve outcomes, and missed requirements — not syntax errors — were identified as the primary failure mode across all tested models.

0
ProgrammingDEV Community ·

Ollama vs LM Studio 2026: Which Local AI Runtime Should You Choose?

A September 2026 comparison of Ollama and LM Studio finds both tools wrap the same llama.cpp engine and cost nothing to start, but differ significantly in design philosophy. Ollama is fully MIT-licensed, CLI and API-first, and best suited for developers building agents, coding assistants, or server-side automation. LM Studio, developed by Element Labs, offers a polished desktop interface with a model browser and chat UI, making it more accessible to non-technical users; it became free for internal business use in July 2025, though the desktop app remains closed source. Both tools now support Apple MLX acceleration, meaning hardware and workflow needs — rather than raw performance — are the deciding factors. Ollama also offers optional paid cloud tiers starting at $20 per month, while LM Studio currently sells no paid plans.

0
ProgrammingDEV Community ·

How to Track Referral Sources Without Storing Raw URLs or UTM Strings

A developer working on RetroPrompt, a web tool for generating portraits, has shared a privacy-conscious approach to referral attribution that avoids storing raw referrer URLs or arbitrary UTM parameters. The method classifies incoming traffic into fixed source categories — such as known search engines, publishing platforms, or an 'other_referral' bucket — by parsing hostnames against an explicit allowlist rather than using substring matching. To preserve attribution across a user session without tracking individuals, the implementation ties entry labels to a short-lived, random tab-session identifier that expires after 30 minutes of inactivity. The author notes that this approach has clear limits: two open tabs can generate separate sessions, and the metric is not equivalent to unique visitors or long-term retention figures reported by other analytics tools. The article emphasizes that attribution labels serve as bookkeeping aids, not proof of causation, and that missing or unrecognized attribution should remain visible rather than be silently discarded.

0
ProgrammingDEV Community ·

Developer Builds AICore, a Rust-Based Adaptive Resource Control Layer for PCs

A developer has created AICore, an experimental computer control layer written primarily in Rust that runs on both Windows and Linux. The project sits between applications and operating-system hardware interfaces, continuously observing workloads and adapting system resources such as CPU, GPU, NPU, and power states accordingly. Unlike static performance modes, AICore follows a cycle of observing, predicting, deciding, executing, and learning from system telemetry and workload context. A key design principle is that any AI inference component can only propose resource policies, not enforce them — deterministic controls remain active if an AI model is unavailable. The system also separates slow AI reasoning from a fast control path to prevent stale or outdated decisions from being applied to rapidly changing system states.

0
ProgrammingDEV Community ·

PostgreSQL Polymorphic Associations: Why One Column Pair Creates Data Integrity Risks

Polymorphic associations, common in frameworks like Rails, Laravel, and Django, store a parent type name and ID in two columns on a child table without any real foreign key constraint. PostgreSQL cannot enforce referential integrity across multiple parent tables from a single column, meaning orphaned records, invalid IDs, and mistyped class names can accumulate silently. When the set of parent tables is small and fixed, the recommended approach is one nullable foreign key per parent combined with a CHECK constraint ensuring exactly one is populated. For larger or more varied parent sets, a shared supertype table restores proper database-level enforcement. The polymorphic column pair should only be retained when the list of parent entities is genuinely open-ended, and its virtual relationships should be explicitly documented in schema diagrams.

0
ProgrammingDEV Community ·

NocoBase v2.1.39 to v2.2.14 upgrade adds 4 tables, 44 columns, drops nothing

A developer tested a NocoBase upgrade from version 2.1.39 to 2.2.14 using official Docker images with PostgreSQL 16 to answer a community concern about what upgrades actually change. The upgrade added four new database tables related to UI layouts, AI usage, and workflow task tracking, while adding 44 new columns — 42 of which belong to those new tables. No existing tables or columns were dropped, and only two columns were added to existing tables. A test page built through the admin UI rendered identically after the upgrade, with the same data, column order, and row count. The investigation also revealed that pages built via the admin UI are stored in a 'flowModels' table rather than the expected 'uiSchemas' table, a detail absent from the official changelog.

0
ProgrammingDEV Community ·

Developer Finds AI Summaries and Shifting Trends Stifle New Open-Source Library

A developer began building a React component library for Bulma v1 in 2024 after finding no up-to-date packages available, spending years writing documentation, blog posts, and SEO-optimized content to attract users. Despite expanding the library well beyond basic wrappers — adding carousels, dialogs, toasts, and a full forms overhaul by mid-2026 — the project accumulated only eleven GitHub stars. The developer attributes the lack of traction to two possible factors: declining general interest in the Bulma CSS framework, and a broader shift toward AI-assisted development tools. A July 2025 Pew Research Center study supports the latter concern, finding that Google search pages displaying AI summaries saw users click through to regular results only 8% of the time, compared to 15% without such summaries. The developer concludes that traditional discovery methods — searching, clicking, and reading library documentation — may no longer reflect how developers choose tools in an AI-driven landscape.

0
ProgrammingDEV Community ·

Developer builds 5 lightweight Python CLI tools to automate CSV cleanup

A developer frustrated with repetitive data-cleaning tasks built five small command-line tools in Python 3.8+ to handle common CSV problems. The tools cover deduplication, whitespace trimming, header normalization, file splitting, and merging, as well as CSV-to-JSON conversion. Each tool performs a single task and prints a summary of changes made, so users can verify the output. The toolkit has no external dependencies and includes a file organizer that sorts folders by type, extension, or date with a dry-run preview mode. The developer plans to package all five tools into a downloadable toolkit and is inviting community feedback via comments.

0
ProgrammingDEV Community ·

Developer Builds Open-Source Gate to Block Unvetted AI Agents from Reaching Production

A developer has released HivePlane, an open-source control plane designed to prevent AI agents from accessing production environments without first passing reproducible benchmarks and holding a signed attestation. The tool was created after years of qualifying agents solely through live demos, a practice the developer found widespread across the industry. HivePlane enforces a certification ladder with distinct thresholds for staging and production, and any change to a prompt, model, or toolset invalidates prior certification. A field test run last week against real agents — including a Python support agent and a LangGraph judge graph — recorded a 10/10 pass rate across 20 acceptance criteria. The system also demonstrated enforcement in practice, issuing 403 errors to uncertified agents, blocking unauthorized model swaps, and terminating runs that exceeded budget limits.

0
ProgrammingDEV Community ·

Metabase SQL injection flaw CVE-2026-72898 grants admin access without login

A critical unauthenticated SQL injection vulnerability, CVE-2026-72898, was discovered in Metabase, a widely used business intelligence platform that stores database credentials for connected data warehouses. The flaw exists in the password-reset endpoint's user-id parameter, which passes unsanitised input into a query, allowing attackers to forge an administrator session via a crafted SQL payload without needing any credentials. CISA added the vulnerability to its Known Exploited Vulnerabilities catalog on 11 August 2026, setting a federal remediation deadline of just three days. The threat group ShinyHunters has been linked to active exploitation, with affected organisations including Mathspace, which reported over one million students and staff impacted across Australia and New Zealand. Patched versions are available across six Metabase release lines from 0.58 through 0.63, and security experts stress that patching must be followed by credential rotation to fully contain any breach.

← NewerPage 942 of 3187Older →