Show HN: Judgement – Multiplayer online card game with hidden easter eggs (playjudgement.vercel.app)

I'm Alex, a tech lead of the Selenium project and a QA engineer. For the last 10 months, I've been working on Alumnium - a test automation library that should make my life easier and let me write less code when automating the tests. The goal is to use LLM for _automating_ so I could focus more on _testing_.

Unlike other similar projects, Alumnium can be used in existing tests without changes to test runners, reporting tools, or any other test infrastructure. This allows me to gradually migrate my test suites (mostly Selenium) and revert whenever something goes wrong (this happens a lot, to be honest).

Other major differences:

- cheap (works on low-tier models like gpt-4o-mini, costs $20 per month for 1k+ tests)

- not an AI agent (dumb enough to fail the test rather than working around to make it pass)

- fast thanks to a built-in caching

- supports both mobile (Appium) and web (Playwright, Selenium)

- supports completely local execution (Ollama)

The project is still in a very much experimental phase, but I'd be happy to get any feedback.

Comments (2)

pydry · 1d ago

This looks like a recipe for making test flakiness even worse.

p0deje · 1d ago

It's quite stable on simple applications, but the flakiness definitely increases with more complex apps. I think there is a lot to do to fix it, one idea is to trim the accessibility tree to a narrow region that is relevant, then use it in a prompt. For example, if you want to "sort by price", the LLM doesn't really need to know about anything beyond the filters region of the app.

We still can't stop plagiarism in undergraduate computer science (2018) (kevinchen.co)

Catching the Silent Threat: How Dynamic Analysis Revealed an NPM Attack Chain (safedep.io)

Peeking Behind the Code–IRS Just Open-Sourced Direct File (forbes.com)

My front row seat on 'Inside the NBA' (nytimes.com)

Gene-edited pig kidneys moving long-stymied field of xenotransplantation forward (science.org)

Google Photos API changes from March 2025 (2024) (github.com)

Google as you know it is slowly dying (vox.com)

PURL Administration (purl.archive.org)

Show HN: Judgement – Multiplayer online card game with hidden easter eggs (playjudgement.vercel.app)

Confidence Unlocked: A Method to Measure Certainty in LLM Outputs (medium.com)

More than half of top mental health TikToks contain misinformation (theguardian.com)

SyntheMol-RL: reinforcement learning framework for designing novel antibiotics (biorxiv.org)

Twin modelling reveals partly distinct genetic pathways to music enjoyment (nature.com)

China Plans Major Nuclear Expansion with 10 New Reactors (oilprice.com)

Unexpected Gotchas in Making a Game Deterministic (jfgeyelin.com)

QuantumAccel: A High Performance Quantum-Inspired Logic Library in Rust+Python (github.com)

Why don't nearly half of Americans have any investments? (investmentnews.com)

Ask HN: Would you use a platform that deploys Docker Compose apps?

Copy from a Simpler Language (tidyfirst.substack.com)

Golden Algebra: A unifying mathematical framework (github.com)

Me at the Zoo (en.wikipedia.org)

Google released a mobile app that lets you download and run AI models locally (techcrunch.com)

LLM Exposure (thelastwave.substack.com)

Show HN: A site for YC rejection stories (ycrejection.com)

MIT bans class president from graduation commencement after pro-Palestine speech (nbcnews.com)

S&P 500 Annual Returns (fool.com)

NFT Marketplaces Attempt to Regain Relevance with Token Trading (thedefiant.io)

Intelligent Agent Technology: Open Sesame! (1993) (blog.gingerbeardman.com)

A good coding bootcamp question

Unit Price Shopper [video] (youtube.com)

Do night owls decline cognitively faster than early birds? (umcgresearch.org)

We're beating $359M in funding with two people and OCaml (terrateam.io)

Prime numbers have fascinated for centuries−tech revolutionizing search for them (theconversation.com)

Yandex Releases Largest Event Dataset for Advancing Recommender Systems (bigdatawire.com)

Qatar Museums brings over 1000 curated masterpieces to Online Collection (thepeninsulaqatar.com)

Software Development Job Postings on Indeed in the United States (fred.stlouisfed.org)

All hail the ELBO: Why you should care about the evidence lower bound (dominiccoey.github.io)

Ukraine's AI-powered 'mother drone' sees first combat use (kyivindependent.com)

Why Meta is in trouble in Nigeria (theconversation.com)

App computes a baseball lineup with pitching and bench for Juan Soto's salary (soto-tracker.streamlit.app)

Close to the metal: web design and the browser (goodinternetmagazine.com)

What's the fuss about MCP? (code examples) (dylanhuang.com)

Why am I filled with nostalgia for a pre-internet age I never knew? (theguardian.com)

Startup wants to make more climate-friendly metal in the US (technologyreview.com)

Why we can't have nice private TLD DNSSEC (egbert.net)

China surrounded Xi aide with seasoned experts to seal U.S. deal (japantimes.co.jp)

Twain Dreams: The Enigma of Samuel Clemens (harpers.org)

North Korean phone has both censorship and accountability in the user interface (alecmuffett.com)

Ask HN: People using AI daily to write – what's the biggest gap you feel?

The source code of the CIH (Chernobyl) virus (github.com)

Show HN: Open-source LLM-powered test automation library for mobile and web

Comments (2)