How Wikidata-based daily trivia generation hit labelling, timeout and data-quality problems

·1 views

A team building daily trivia quizzes sourced every question from Wikidata facts so that answers could be checked and linked back to their source items. They found that many proper names exist only under a multilingual label, so queries filtering by English or French silently dropped items such as Juneau and several famous people. Heavy SPARQL queries also timed out, which they addressed by starting from Wikipedia-mapped item lists in batches and saving results incrementally. Other fixes covered handling multiple conflicting values per property, excluding misleading discoverer and music entries, and correcting outdated names and duplicate continent entries. The article, measured on 23 October 2026, also notes Wikimedia Commons thumbnail sizes that the service accepts.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

GitHub Rebuilds Git Storage as AI Agents Drive Surging Commit and Push Volumes

GitHub has published a blog post describing a rebuild of the storage layer beneath all its Git repositories, driven by the growth of AI agents. In September, developers and agents made 7.38 billion commits, more than five times the previous year, with pushes reaching 3.35 billion a month. The current Spokes system keeps a full copy of each repository on five file servers and requires a majority to agree on every ref update, so adding read capacity by adding copies slows down writes. GitHub plans to separate data safety from traffic serving, running heavy push steps in parallel so only the small ref-update step is serialized. This matters most for the main branch, where merges compete to update the same ref.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

PayGuard: Open-Source SDK Enforces Spend Limits and Risk Checks for AI Agent Payments

PayGuard is an open-source security SDK developed during a Razorpay hackathon to add financial safeguards to transactions made by AI agents. It acts as an automated guardrail that enforces spending limits, verifies idempotency, and runs risk evaluations before any purchase goes through. Users can define financial policies once and let the SDK apply spend caps and checks without human intervention. The tool is designed to run locally or on self-hosted infrastructure so sensitive transaction data stays off third-party servers, and it works with open-weight local models to avoid per-request costs. Because the code is open source, developers can adjust risk models, swap inference backends, or modify rule sets.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

GoldLift Offers a Free Alternative to Fitbod's Paid Workout Planning

GoldLift is presented as a free alternative to Fitbod, which charges about $15.99 per month or $95.99 per year after a 7-day trial and has no permanent free tier. Both apps generate workouts, but GoldLift builds a structured multi-week programme and adjusts per-set loads based on your results, while Fitbod creates each session from muscle freshness and available equipment. The developer, who is affiliated with GoldLift, adds that GoldLift includes an in-workout chat coach, nutrition logging, gamification and social features that Fitbod lacks. The comparison also notes Fitbod's advantages, including its mature recovery modelling, exercise videos and existing Android app, whereas GoldLift currently runs only on iOS. Prices were checked on 23 October 2026 in the US App Store and may differ by region or legacy plan.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Designing a Progress-Aware Controller to Decide When AI Agents Should Stop

A developer argues that deciding when an AI agent should stop is a system-level design problem rather than a matter of prompting or a fixed step cap. Fixed limits like max_steps are useful as safety boundaries but cannot tell whether a task is actually complete, since simple tasks and complex ones need very different numbers of turns. The proposed architecture separates the agent from a controller that checks actions against a budget gate, evaluates progress after each tool call, and tracks completion, progress, and remaining budget. The controller would return one of four outcomes: finish, continue, replan, or stop as incomplete, keeping budget exhaustion distinct from genuine success. The author favors a transparent, rule-based policy for an MVP, durable state storage, and head-to-head testing of fixed-limit, stagnation-aware, and progress-aware policies on success rate, cost, and premature stops.

0 comments Read more at DEV Community