SShortSingh.

Programming

0
ProgrammingDEV Community ·

Dev Builds Theatre Prop Continuity Tracker to Catch Staging Errors

A developer created Prop Continuity Desk, a rehearsal planning web app designed to help small theatre productions track the physical movement of props across scenes. The app is built around a fictional six-scene play with twelve props and detects two key mistakes: placing the same prop in two locations within one scene, and moving a prop between scenes without a recorded transfer cue. It uses Next.js, React, TypeScript, and Sanity as its content backend, with scene reviews stored in separate documents to avoid overwriting the data being reviewed. A practice-copy feature lets visitors explore the rules without altering the live production plan, and any exported cue sheet from a practice session is clearly labelled as such. The project was built using Codex starting 26 September 2026, with the AI proposing the scenario, writing the code, and running the tests under the account owner's authorization.

0
ProgrammingDEV Community ·

Sekura JS Explores Sharing Software Verification Proofs Without Exposing Source Code

Sekura JS is developing an approach that separates a software module's source code from its verification contract (SJV) and proof artifact (SJP), allowing developers to share evidence of formal verification without disclosing proprietary implementations. The SJV document specifies the properties a module claims to satisfy, while the SJP carries the mathematical evidence produced during verification. A third-party verifier can replay the stored SMT obligations using tools like Z3 or CVC5 to confirm their validity without ever seeing the underlying source code. However, the project acknowledges a key limitation: replaying an SJP can confirm that the mathematical proofs are valid, but cannot independently verify that those proofs were actually generated from the specific private codebase the developer claims. This creates an explicit trust boundary between the private source code and the proof artifacts exchanged between parties.

0
ProgrammingDEV Community ·

Where Game Development Budgets Actually Go, Explained by Category

Game development costs are typically split across five core categories: engineering, art and audio, design, QA and testing, and project management. Engineering tends to consume the largest share, usually between 40 and 50 percent, covering gameplay systems, networking, platform work, and performance optimisation. Art and audio follow at 20 to 30 percent, while design, QA, and project management each account for roughly 5 to 15 percent. These ranges are based on agency-scoped projects and shift depending on genre, platform targets, team structure, and scope. When budgets are limited, understanding which category drives the most complexity for a given project can help teams prioritise spending more effectively.

0
ProgrammingDEV Community ·

Why Organisations Must Modernise Ageing Educational Games Before They Vanish

Thousands of educational games built between 2000 and 2015 are becoming inaccessible due to obsolete technologies such as Adobe Flash, discontinued in December 2020, and legacy platforms like Java applets and early Unity builds. The shift in classroom devices toward Chromebooks, iPads, and smartphones has further rendered older games unusable, as they were designed for desktop monitors with mouse-and-keyboard input. Accessibility gaps are another pressing concern, with many legacy games failing to meet WCAG 2.2 standards published in 2023, leaving learners with disabilities unable to engage with the content. Institutions in the UK, US, and elsewhere also face legal obligations under laws such as the Equality Act 2010 and the ADA, making non-compliant content a potential liability. Modernising these tools is presented not as a discretionary upgrade but as a necessary step to preserve past educational investment and meet current technical, legal, and pedagogical expectations.

0
ProgrammingDEV Community ·

Technical Guide: Building Cross-Platform Cloud Save Systems for Unity Games

Game developers at ShortSingh outline a technical framework for building robust cloud save systems in Unity, addressing challenges that arise when games span multiple platforms and devices. The guide draws on real-world experience from projects including Domi Online, a persistent MMORPG, and RuneScape Mobile developed at Jagex. Key engineering problems covered include conflict resolution when two devices hold divergent saves, offline-first design for players with intermittent connectivity, and schema migration when game updates change save data formats. The recommended architecture separates save logic into three distinct layers — data model, local storage, and cloud sync — to keep each component independently testable and replaceable. Platform-specific considerations for Google Play Games Services, Apple Game Center, and Steam Cloud are also addressed, as each service carries its own failure modes and rate limits.

0
ProgrammingDEV Community ·

AI-Assisted Backend Dev: Why Compiling Code Is Not Enough to Confirm It Works

A developer building FinanceHub's custom backend used an AI agent to write code task-by-task, guided by a strict contract-first approach where the OpenAPI spec and database schema were treated as fixed inputs. Despite code compiling cleanly and AI explanations sounding plausible, a mandatory 6-step verification process per task caught several critical bugs that automated tests missed. Key failures included Supabase JWTs being silently rejected due to an algorithm mismatch, a PATCH endpoint overwriting optional fields with null on partial updates, and 24 row-level security policies that were effectively dead code because the required schema-level GRANT permissions were never set. Each bug was only discovered because verification demanded real tokens, real partial payloads, and actual database permission checks — not synthetic stubs. The experience highlights that with AI-generated code, the central question shifts from 'does it work?' to 'how do I actually know it works?'

0
ProgrammingDEV Community ·

How to Build a Fast Local Search Index from Your X (Twitter) Archive

A developer has outlined a method to convert a large X (formerly Twitter) data archive — typically 150 to 400 MB — into a lightweight, queryable local search index. The core challenge is that parsing the archive's main file, tweets.js, as a single JSON object causes Node.js to exceed its default memory heap limit. The solution involves streaming the file and processing one record at a time, so memory usage stays proportional to a single post rather than the entire dataset. The resulting index, stored as NDJSON, can be searched instantly using grep or loaded into DuckDB for ranked, full-text queries. Because post IDs increase monotonically, the index can be updated incrementally when a new archive export is downloaded, avoiding a full rebuild.

0
ProgrammingDEV Community ·

Adobe Patches Critical 9.9 SQL Injection in Connect 12.12 Alongside Eight Other Flaws

Adobe released a September 2026 security update for Adobe Connect that addresses nine vulnerabilities, seven of which are rated critical. The most severe flaw, CVE-2026-75682, carries a CVSS score of 9.9 and is a SQL injection vulnerability that can allow a low-privileged attacker to execute arbitrary code. Other critical issues include stored cross-site scripting bugs that enable privilege escalation and a path traversal flaw requiring no authentication or user interaction. The fixes apply to Adobe Connect 12.12 and Adobe Connect Android Mobile App 4.5, with earlier versions remaining vulnerable. Adobe stated it is not aware of any active exploitation of these vulnerabilities at the time of the advisory.

0
ProgrammingDEV Community ·

GitHub's pull_request_target changes will break workflows in 27% of top repos by November

GitHub is rolling out two breaking changes to its pull_request_target Actions trigger, with a key deadline on November 2, 2026, when the trigger will be blocked on public repositories lacking an explicit Actions policy. A scan of the 1,000 most-starred GitHub repositories conducted on September 26, 2026 found that 269 of them run at least one workflow using the affected trigger, putting those workflows at risk of silently stopping. Since July 20, 2026, the actions/checkout action also refuses to check out fork pull request code in privileged workflows unless maintainers explicitly opt in. More critically, three repositories were found using an older, unpatched version of actions/checkout that bypasses the new guard, potentially exposing workflow tokens and secrets to fork code. A free open-source tool called prt-check has been released to help repository maintainers audit their own workflows for these vulnerabilities.

0
ProgrammingDEV Community ·

Developer shares human-led AI workflow to build apps from idea to working API

A developer has published a detailed guide demonstrating how to build software applications using AI tools without producing low-quality, unreliable output. The workflow was tested through a demo project called Orbit, a Slack-style team chat app, built with assistance from ChatGPT, Claude, and Grok. The process involved setting clear requirements, reviewing AI-generated changes at each stage, and deriving an API only after the user interface behavior was accepted. The author emphasizes that AI cannot self-certify code as production-ready, and that human oversight at every handoff is essential to useful results. Orbit itself remains a limited demo with no authentication, shared public data, and no persistent storage across restarts.

0
ProgrammingDEV Community ·

Developer Builds AI Smart TV OS on AWS After Tackling Vega OS Compatibility Hurdles

A developer building Vid30, a self-described AI Streaming Operating System for Smart TVs, documented the first sprint of their public build journey on DEV Community. Despite warnings that Vega OS development is designed for Linux and macOS environments, they chose to proceed using an Ubuntu EC2 instance on AWS as a remote development environment. Using Kiro as an AI-assisted IDE, they successfully forked Amazon's React Native Multi-TV starter project and learned the Vega SDK from scratch, encountering multiple issues along the way including Git, Yarn, Metro, and SDK compatibility problems. The developer resolved most obstacles but hit a final blocker in the Vega simulator workflow, publicly asking the Fire TV and Vega developer community for help. They are also seeking a physical Fire TV Stick from Amazon to test Vid30 on real hardware, framing the experience as a lesson in knowing when to pivot rather than pursuing every problem to completion.

0
ProgrammingDEV Community ·

Engineer's Go/No-Go Rubric for Evaluating MCP Servers Before Production

A structured two-tier rubric has been proposed for engineers to objectively decide whether an MCP server is ready for production, moving beyond generic checklists. The framework distinguishes between Tier 1, which covers protocol and handler correctness, and Tier 2, which assesses real-world agent-facing behavior during actual tasks. A key insight is that passing all protocol and unit tests does not guarantee an MCP server will function correctly when a non-deterministic agent is selecting tools, constructing arguments, or recovering from errors. Common gaps identified include front-loaded authentication checks, undetected partial failures, and schema drift between documented inputs and runtime handler behavior. Engineers are advised to achieve a minimum score of five out of six Tier 1 checks before proceeding to agent-level evaluation.

0
ProgrammingDEV Community ·

System Prompts Are Not Guardrails: Why AI Agent Safety Needs External Controls

Developers commonly place safety rules for coding agents inside system prompts, but this approach offers weaker protection than it appears. System prompts share the same channel as user messages and tool outputs, making policy rules vulnerable to being overridden, forgotten, or reinterpreted by the model. Unlike external controls, prompt-based rules are difficult to version, test on a subset of agents, or roll back quickly when something goes wrong. They also leave poor audit trails, making it hard to diagnose exactly which control failed after an incident. Experts recommend moving hard limits into external enforcement layers — such as tool allowlists, scoped credentials, and gated approval steps — while reserving the system prompt for style and reasoning guidance only.

0
ProgrammingDEV Community ·

Diagrams as Code: How to Keep Architecture Docs in Sync with Your Codebase

Outdated architecture diagrams stored as PNG or draw.io files are a common pain point for engineering teams, as they cannot be diffed, reviewed, or reliably updated alongside code changes. The 'diagrams as code' approach treats diagrams as plain text files, making them trackable via Git, reviewable in pull requests, and verifiable by CI pipelines. Tools like Mermaid, PlantUML, D2, and Structurizr DSL offer varying strengths, with Mermaid being a low-friction default since GitHub and GitLab render it natively. A practical workflow involves requiring any architecture-changing PR to include a corresponding diagram update, enforced through PR templates and automated syntax validation via mermaid-cli. Architecture data already present in repos — such as docker-compose or Terraform files — can also be parsed programmatically to auto-generate diagrams, reducing the risk of them becoming stale.

0
ProgrammingDEV Community ·

AI-Built Tool Checks Raspberry Pi HAT Pin Conflicts Before You Stack Boards

A developer named Anurag used the AI coding agent Claude Code to autonomously build 'Will It Stack', a web tool that determines whether Raspberry Pi add-on boards (HATs) can safely share a 40-pin GPIO header. The tool cross-references structured pinout data from pinout.xyz to detect hardware conflicts such as shared GPIO pins or duplicate I2C addresses, rather than relying on generative text responses. For example, it identifies that a Fan SHIM and Inky pHAT both claim GPIO 17, while an ADC Pi and RTC Pi can coexist if one chip's I2C address is remapped. The checker runs deterministically without consuming AI model quota, and the live agent interface is available free at will-it-stack.vercel.app with no login required. The project's source code has been released on GitHub under the MIT license, with imported hardware data carrying a CC BY-SA 4.0 license.

0
ProgrammingDEV Community ·

AI Startup Jev Claims 444x Cost Savings, but Independent Proof Remains Absent

A startup called Jev emerged from stealth on September 15 with $40 million in funding, claiming its AI delivers decisions in milliseconds at 444 times lower cost than competing models. Unlike conventional AI, Jev returns typed answers with confidence scores rather than generated text, positioning itself as a lightweight decision-making API. The launch drew immediate attention from developers, who overwhelmed the API on day one, while the tech community remained divided between early adopters and skeptics demanding independent benchmarks. A detailed audit of the launch materials found that while Jev's architectural approach and latency profile appear plausible, the specific 444x cost figure lacks verifiable third-party evidence. The episode highlights a recurring issue in AI product launches, where headline performance claims often rest on narrow or undisclosed testing conditions buried in fine print.

0
ProgrammingDEV Community ·

Developer Rejected 2 Minutes After Applying, Blaming Automated Degree Filter

A developer applying for a junior solutions engineer role received an automated rejection roughly two minutes after submitting their application. Despite lacking a STEM degree — listed as a requirement — the applicant had hands-on experience building a custom inventory management system integrated with WooCommerce for a real business client. The project involved API integrations, database work, multi-device synchronization, and deployment, going well beyond portfolio or tutorial work. The near-instant turnaround strongly suggests an automated applicant tracking system filtered out the candidate before any human reviewed their tailored resume, cover letter, or written responses. The incident reignites debate over whether 'skills-based hiring' rhetoric in tech is undermined when degree requirements remain as hard automated filters at the top of the pipeline.

0
ProgrammingDEV Community ·

Google Gemma 4 26B Runs on One TPU v6e with 15x More KV Cache via QAT Build

A developer has published a step-by-step guide to serving Google's quantization-aware-trained (QAT) Gemma 4 26B-A4B model on a single Google Cloud TPU v6e chip using vLLM. The QAT build uses only 17.43 GiB of HBM and delivers 53,888 tokens of KV cache and 1,283 output tokens per second, compared to RedHat's FP8 build which uses 27.99 GiB and yields just 3,456 tokens of KV cache at 668 tokens per second. The improvement was achieved by repacking Google's own unquantized QAT weights into a W4A16 Compressed Tensors format, filling a gap left by Google's official vLLM support, which excludes the 26B model size. Despite the large efficiency gains, both builds scored within one percentage point of each other on a 3,880-record classification benchmark. The resulting checkpoint has been released publicly on Hugging Face, and the repacked model also loads on NVIDIA L4 GPUs without additional patches.

0
ProgrammingDEV Community ·

How a Brazilian startup built fan-out on BullMQ without adding a message broker

The CTO of Monest, a Brazilian debt-collection startup running a NestJS backend, faced a fan-out problem when multiple teams and client integrations all needed to react to the same 'tag applied' event in real time. BullMQ, their existing job queue, is designed for one-event-one-consumer processing and does not natively support delivering a single event to multiple listeners. Rather than introducing a dedicated message broker like RabbitMQ or Kafka — which would add infrastructure, operational overhead, and on-call complexity for a lean team — the team evaluated what BullMQ was actually failing at and found it was only one narrow gap. Their solution was to keep a single tag-applied worker that sequentially calls all downstream handlers, effectively achieving fan-out without any new infrastructure.

0
ProgrammingDEV Community ·

Developer Replaces Failed GitHub Activity Graph With Self-Hosted Static Solution

A developer's GitHub profile README activity graph went offline after the third-party Vercel service hosting it returned an HTTP 402 error and paused its deployment. The graph had been generated on demand by a public external service, creating a dependency outside the developer's control. Switching to another hosted URL would have simply replaced one unreliable dependency with another. To fix this, the developer built a small JavaScript generator paired with a GitHub Actions workflow that regenerates and stores the graph image directly in their own repository. The graph now updates automatically once a day and can also be refreshed manually, eliminating reliance on any external hosting service.

← NewerPage 885 of 3177Older →