SShortSingh.

Programming

0
ProgrammingDEV Community ·

Bug Let a Verifier Confirm '13 of 13' Completeness Using Just One Cell

A developer discovered that their claim-verification system could incorrectly resolve completeness claims like '13/13' by finding a single cell containing the number 13, using it as evidence for both sides of the pair. Because the checker matched each side of a fraction independently, equal numerator-denominator pairs required only one matching value instead of two distinct cells. A one-line code fix enforced that the right-hand side must resolve to a different cell than the left, which immediately caused three previously passing claims to show as unresolved. The developer also identified a separate, subtler issue: a carried-over table was visible in the document but never read by the counter, meaning its pairs produced no findings at all — neither passing nor failing. The key takeaway is that a verifier made stricter should always cause some failures to surface; if it does not, the fix either missed its target or no reachable claims were affected.

0
ProgrammingDEV Community ·

Three MCP Servers That Extend Claude Desktop With Files, GitHub, and Postgres

A developer has shared three Model Context Protocol (MCP) servers they actively use with Claude Desktop to streamline daily workflows. The filesystem server allows Claude to directly read project files, eliminating the need to manually copy and paste code into the chat. The GitHub server enables Claude to review pull requests, check issues, and browse repositories without leaving the conversation. A PostgreSQL server grants Claude direct database query access, useful for tasks like identifying user activity patterns. Each server can be installed via the mcp-hub CLI in roughly two minutes, though users are cautioned to restrict filesystem path access carefully to avoid unintended disk exposure.

0
ProgrammingDEV Community ·

Mainframe Migration Tools: What They Do Well and Where They Fall Short

Mainframe migration tools fall into four broad categories: discovery and analysis, rehosting and emulation, automated code translation, and data migration — and most real programmes require tools from at least three of them. Vendor demos of automated COBOL-to-Java conversion often look impressive, but are typically run on cleaner code than what organisations actually have in production. While automated translation reliably produces runnable code, it rarely produces maintainable code, and bridging that gap is where the majority of migration budgets are spent. Discovery and analysis tools are considered the most consistently valuable category, as no organisation typically has a full picture of a system built up over decades. Buyers are also advised to scrutinise vendor ownership before signing long-term agreements, as consolidation through acquisition is common in this market.

0
ProgrammingDEV Community ·

Developer admits mislabeling a build failure as 'unknown' to avoid a hard negative result

A developer building a check system that uses three result states — pass, fail, and unknown — incorrectly marked a detected dependency cycle as 'unknown' instead of 'fail', a mistake caught by a teammate. The author argues the 'unknown' state should only apply when a check cannot reach a conclusion, such as when an input file is missing or a human-produced record cannot be re-run. A completed graph traversal that confirms a cycle is a definitive negative answer, making it a clear failure. The error was compounded by the fact that the 'unknown' label felt principled and humble, while quietly removing a red flag from the build summary. The author's key takeaway is that if a check ran to completion and returned a negative result, it must be recorded as a failure — 'unknown' is for the absence of an answer, not for an answer that is inconvenient.

0
ProgrammingDEV Community ·

Why Treating Redis as a Database Can Bring Down Your Entire Backend

Redis is designed as a cache — a store of recreatable data — not an authoritative database, and conflating the two roles introduces serious architectural risks. When Redis goes down, a naive fallback to the primary database can trigger a failure cascade, as the database suddenly absorbs traffic it was never sized to handle alone. Techniques such as deliberate timeouts prevent slow Redis responses from tying up application resources indefinitely, while circuit breakers stop the system from repeatedly hammering a known-unhealthy dependency. Uncontrolled retries can worsen an outage by multiplying request volume against an already struggling service. Building resilient backends requires treating Redis failures as expected architectural events and designing explicit, measured responses to each failure mode.

0
ProgrammingDEV Community ·

Single faulty server caused hours of checkout failures while going undetected

A degraded network interface on one of twenty-four server instances caused checkout error rates between 3–8% for most of a Thursday, going unnoticed because fleet-wide monitoring diluted the signal. The instance continued passing health checks by successfully responding to lightweight synthetic requests, even as roughly one in ten real connections stalled and timed out. It took three hours to isolate the culprit, despite the cloud provider having flagged a hardware degradation notice ninety minutes before the first customer complaint. In response, the team introduced per-instance error rate alerts, passive load-balancer health checks based on real traffic, and a pipeline for hardware health events. The key operational change was to remove a suspected instance immediately and investigate afterward, rather than spending time confirming the obvious.

0
ProgrammingDEV Community ·

How ~900 Silent Catch Blocks Were Hiding Real Payment Failures in Frontend Code

A frontend engineer discovered that HTTP 200 responses from a payment BFF were masking real user-facing errors, including TypeErrors on payment buttons, across four separate frontend repositories. A static grep of the codebases revealed approximately 900 empty or swallowed catch blocks — none of which produced runtime telemetry. The engineer rejected default SDK configurations, opting instead for conservative sampling rates and mandatory domain tags to comply with Brazil's LGPD data privacy rules and avoid PII leakage to third-party vendors. Rather than replacing existing Grafana and APM tooling, the approach was designed to fill a specific gap: capturing client-side errors tied to release versions and user journeys. The resulting setup treats frontend observability as a revenue-gap problem, not a tooling exercise, with sampling rates and tagging contracts documented before any SDK was deployed to production.

0
ProgrammingDEV Community ·

Browser Experiment Quantifies Chaos: Double Pendulum Diverges in Under 8 Seconds

A developer used browser-based physics simulators to numerically measure chaotic behaviour in two classic systems: a double pendulum and the logistic map. Two simulated double pendulums released from positions just 0.057 degrees apart remained synchronised for roughly 5.6 seconds before fully diverging by 7.2 seconds. The calculated Lyapunov exponent of approximately 1.095 per second confirms exponential error growth, meaning any initial uncertainty doubles repeatedly until long-term prediction becomes impossible. The logistic map independently demonstrated the same universal route into chaos through period-doubling, with the ratio between successive doublings converging on the Feigenbaum constant of 4.669. Both systems are fully deterministic, highlighting that chaos arises not from randomness but from extreme sensitivity to initial conditions.

0
ProgrammingDEV Community ·

Developer builds self-hosted AI running coach that syncs with Garmin and plans workouts

A software developer has built a self-hosted AI coaching agent that automatically reads data from Garmin and Strava — including sleep, HRV, and lap splits — to plan daily running and strength sessions. The system, running on a Mac mini using an agent framework called Hermes, messages the user each morning, writes workouts directly to Garmin Connect, and grades completed sessions against its own prescriptions. It also tracks fitness trends over time, advises on nutrition around hard efforts, and adjusts training load based on recent activity — for instance, replacing a time trial with an easier session after a heavy previous day. The developer credits the setup with helping them achieve a sub-20-minute 5K goal and saving roughly an hour of weekly training planning. A stripped-down version of the configuration has been published on GitHub for others to adapt.

0
ProgrammingDEV Community ·

Senior Backend Engineer Pivots to AI Infrastructure After Copilot Narrows Experience Gap

A senior backend engineer with four years of experience in Java, Spring Boot, and distributed systems noticed that AI coding tools like GitHub Copilot were allowing junior engineers to match his output on routine tasks. Rather than fearing direct replacement, he observed that the skill gap between senior and junior developers was compressing as AI automated boilerplate coding and pattern recognition. He shifted his focus away from conventional backend work and toward AI backend engineering — the discipline of building retrieval pipelines, agent loops, tool-call protocols, and guardrails that power production AI systems. He argues that this role sits at a narrow intersection between traditional backend engineering and AI infrastructure, a space where few professionals currently have both skill sets. Working full-time, he spent several months building out this expertise, positioning himself for a role he believes will be in high demand by 2026.

0
ProgrammingDEV Community ·

Assigning End-to-End Process Owners Solved IT's Cross-System Accountability Gap

A technology team discovered that a detailed application register with 240 entries and named owners failed to prevent an order from taking 11 days to reach a customer. The delay occurred because three systems — order capture, credit checking, and dispatch — each functioned correctly, but no one owned the gaps between them. A rule change in the credit department silently stalled the order, and dispatch never received it since it remained unapproved with no escalation trigger defined. To address this, the team designated business owners for each of 12 end-to-end processes, granting them authority to resolve cross-departmental disputes and approve rule changes affecting other steps. The shift also introduced process-level reporting alongside traditional availability metrics, significantly reducing unresolved inter-departmental conflicts reaching IT leadership.

0
ProgrammingDEV Community ·

DEV Community Thread Asks Members to Share the Best Articles They Have Read

A DEV Community user launched a discussion thread inviting members to share links to the best articles they have recently read, rather than written. The initiative was inspired by a similar thread from user Giorgi Kobaidze, who had asked writers to share their proudest work. The new thread aims to surface high-quality content that rarely trends but holds genuine value for readers. The original poster noted that despite an abundance of good writing on the platform, much of it remains difficult to discover. The thread welcomes all formats, including tutorials, deep dives, and personal tech stories.

0
ProgrammingHacker News ·

UTK COSC562 Course Publishes Device Drivers Lab Exercise Online

The University of Tennessee Knoxville has made a device drivers lab exercise publicly available as part of its COSC562 course curriculum. The material is hosted on the EECS department website under instructor smarz1's course page. The exercise appears to be designed to give students hands-on experience with writing or working with device drivers. The resource was shared on Hacker News, where it received modest engagement with 7 points and one comment.

0
ProgrammingDEV Community ·

SEO Expert Urges Marketers to Query Raw LLM APIs to Expose Retrieval Gaps

Leeds-based SEO consultant Martin McGarry, with two decades of industry experience, recommends a simple terminal experiment to understand how AI chat products differ from raw language models. By querying a bare LLM API directly and comparing the response to the same question in ChatGPT or Gemini, users can observe how retrieval-augmented generation (RAG) fills in the gaps the base model cannot. McGarry argues this demonstrates that SEO is not obsolete, since LLMs must still fetch external content to produce useful answers, meaning website visibility in those retrieval sources still matters. He suggests that rather than focusing solely on llms.txt or markdown files, practitioners may benefit from improving machine-readable outputs like RSS feeds and JSON APIs, which are likely surfaces for server-side fetches. McGarry acknowledges this is a hunch, but believes optimizing existing feed infrastructure could be an overlooked opportunity in the emerging landscape of AI-driven search.

0
ProgrammingDEV Community ·

Silent duplicate IDs caused a correct-looking failure three steps from the source

A work-tracking register allowed a second entry to be created with an already-existing ID, because the uniqueness rule was documented in prose but never enforced in code. When the system processed a landing command, it acted on the older row sharing that ID, leaving the newer one stranded without any error being raised. The fault only surfaced later when an unrelated operation was blocked by a conflict check that was technically accurate against the register but incorrect about real-world project state. The author argues that ID uniqueness must be checked at the point of insertion and across all row states, not just open ones, since state-filtered uniqueness is not true uniqueness. A constraint enforced only by memory and convention will fail silently, and when it does, the visible symptom will point nowhere near the actual cause.

0
ProgrammingHacker News ·

Hacker News Community Shares September 2026 Projects and Curiosities

Hacker News hosts its recurring monthly thread where members share what they are currently building or exploring. The September 2026 edition drew participation from the community with 10 comments and 12 points. The thread serves as an open forum for developers, engineers, and tech enthusiasts to discuss ongoing projects and intellectual interests. It is a long-running community tradition on the platform designed to encourage peer discovery and informal knowledge sharing.

0
ProgrammingDEV Community ·

OpenResearch runs parallel AI research agents locally with full experiment traceability

OpenResearch, an open-source tool developed by the alphaXiv team, allows multiple AI agents to explore different research directions simultaneously using separate git worktrees to avoid file conflicts. The project, built primarily in Rust and licensed under MIT, emphasizes reproducibility by creating immutable, git-native archives of every experimental run so researchers can trace exactly which code version produced which result. An Autoresearch mode enables agents to autonomously propose ideas, modify code, run experiments, and decide next steps without human intervention at each stage. The tool runs entirely on localhost with a local SQLite database, and the developers explicitly state that project creation and experiment runs do not transmit user code externally. Usage telemetry is opt-out and is documented to exclude code, prompts, file contents, tokens, and other sensitive identifiers, with a simple command available to disable it entirely.

← NewerPage 1366 of 3313Older →