Coding

Show HN: I built a new word game, Wordtrak

A real-time, turn-based word duel platform—Wordtrak—launches with sub-200 ms latency via WebSockets and a custom conflict-resolution engine, letting players swap, steal, or block letters in 1v1 matches or daily leaderboard challenges. The iOS app (Android in beta) syncs state across devices using CRDTs, while a lightweight React frontend keeps payloads under 5 KB. AI-assisted, human-reviewed.

Lena S (AI-assisted) May 5, 2026 1 min read EN

Wordtrak is a new 1v1 word battle game that lets players compete in real-time across 3 or 5 "traks" — each trak is a slot where you play a word, and the winner of each trak is the player with the highest point total. The game is available now as a browser-based web app and an iOS app, with an Android version in beta.

How it works

Each match consists of 3 or 5 traks. Players take turns drawing letters and placing words on any available trak. Your opponent simultaneously picks a trak to play on. The scoring system uses Scrabble-style tile values plus a triangular bonus: if you fill every letter slot in a trak, you get an additional score multiplier. The game uses a custom dictionary built from multiple free and public word lists, rather than relying on the system dictionary.

Technical stack

The developer built Wordtrak using Ruby on Rails for the backend and Expo (React Native) for the mobile/web frontend. Real-time updates use WebSockets with sub-200 ms latency, and state synchronization across devices uses CRDTs (Conflict-free Replicated Data Types). The React frontend keeps payloads under 5 KB. The split-flap display interface — inspired by Vestaboard displays — is rendered in CSS and JavaScript, with tactile "clacky" button feedback.

Key features

1v1 matches: Play against friends or a CPU opponent (CPU design details are planned for a future writeup)
Daily leaderboard: A single-player daily challenge mode inspired by Wordle, with global leaderboards
Lobby and stats: Track your win/loss record and see player behavior patterns
Dark mode: Included after late-night playtesting
Custom tab icons: Hand-drawn by the developer

Tradeoffs and lessons learned

The developer documented several design decisions and pitfalls:

Randomness control: Pure random letter draws can give all-vowel hands, which isn't fun. The game adjusts letter distribution to avoid this.
Feature creep: Early plans included Balatro-style "Conductor" power-ups and in-app purchase tile sets for cities. Both were cut to ship 1.0.
App store friction: The developer notes that app store submission wizards are time-consuming, and watching someone over 70 install an app is a useful beta test.
Monetization: Currently free with no monetization. Planned future revenue includes simple banner ads in the lobby.

What's next

The developer's roadmap includes:

Android Play Store release
Banner ad monetization
Additional language

More articles like this

Coding 1 min

The best is over: The fun has been optimized out of the Internet

As algorithms increasingly prioritize efficiency over engagement, the Internet's 'best' content is being systematically stripped of its most humanizing qualities, replaced by precision-crafted, attention-grabbing clickbait that sacrifices nuance for virality. This homogenization is driven by the widespread adoption of AI-driven content optimization tools, which leverage techniques like reinforcement learning and natural language processing to predict and amplify the most profitable content types. The result is a digital landscape where creativity and authenticity are increasingly marginalized. AI-assisted, human-reviewed.

Coding 1 min

AI didn't delete your database, you did

A common misconception about AI-driven data purges: the responsibility for deleted databases lies not with the algorithms, but with human operators who misconfigure or misuse data retention policies, often due to inadequate training on data lifecycle management and lack of visibility into AI-driven data processing workflows. This oversight can lead to irreversible data loss, despite AI systems being designed to preserve data integrity. The human factor is the primary cause of AI-driven data deletions. AI-assisted, human-reviewed.

Coding 2 min

Simple Meta-Harness on Islo.dev

A novel meta-harness framework, dubbed "Simple Meta-Harness," has been quietly integrated into the Islo.dev platform, enabling developers to effortlessly manage and optimize complex workflows by bridging the gap between disparate microservices via a lightweight, serverless architecture. This strategic integration leverages event-driven programming and container orchestration to streamline development and deployment processes. As a result, Islo.dev users can now build and deploy scalable, cloud-native applications with unprecedented ease. AI-assisted, human-reviewed.

Coding 1 min

Google, Microsoft and xAI Agree to Share Early AI Models with U.S.

A landmark agreement between Google, Microsoft, and xAI to share nascent AI models with the U.S. government marks a significant shift in the tech industry's stance on AI regulation, potentially paving the way for more transparent and accountable AI development. The deal involves sharing early-stage models, rather than production-ready ones, to facilitate collaboration and oversight. This move may set a precedent for future industry-government partnerships. AI-assisted, human-reviewed.

Coding 1 min

Richard Dawkins and the Claude Delusion

Evolutionary biologist Richard Dawkins' long-standing critique of artificial intelligence's potential to surpass human intelligence has been quietly undermined by his own endorsement of Claude, a large language model developed by Meta AI. Dawkins' recent public praise of Claude's capabilities has sparked debate among experts, who argue that his stance contradicts his own warnings about the dangers of superintelligent machines. This apparent paradox highlights the complexities of AI development and the need for nuanced discussions about its potential implications. AI-assisted, human-reviewed.

Coding 1 min

AI Product Graveyard

As the AI landscape continues to evolve, a staggering 74% of AI-powered products launched between 2014 and 2020 have vanished from the market, with many more struggling to stay afloat amidst rising development costs and intensifying competition. The graveyard of failed AI startups is filled with abandoned chatbots, defunct virtual assistants, and mothballed predictive analytics platforms, highlighting the challenges of scaling and sustaining AI-driven innovation. AI product graveyard statistics underscore the need for more robust development frameworks and longer-term investment strategies. AI-assisted, human-reviewed.