Alternatives to Show HN: Watch LLMs Play NetHack

I’m experimenting with modeling tabletop RPG adventures as explicit narrative state rather than linear scripts. Everdice is a small web app that tracks conditional scenes and choice-driven state transitions to preserve continuity across long or asynchronous campaigns. The core contribution is explicit narrative state and causality, not automation. The real heavy lifting is happening in the DM Toolkit/Run Sessions area, and integrates CAML (Canonical Adventure Modeling Language) that I developed

Hi HN, I built Savior, a tiny, dependency-free JavaScript library that autosaves form input and restores it after refreshes, tab closes, or browser crashes. It’s fully client-side, framework-agnostic, and designed to handle real-world edge cases like dynamic fields, corrupted storage, and failed submits without crashing the app. Savior restores drafts verbatim — no judgment, no “smart” recovery — just predictable behavior. GitHub: https://github.com/Pepp38/Savior Docs: https://zippers.dev Happy

Hey HN, I'm George. I spent few months rebuilding Obooko, a free ebook platform that's been around since 2010 with 11 million downloads. Reading is losing to scrolling, but BookTok has 370B+ views. People want book content, they're just not reading books. I think that’s because modern reading hasn’t kept up with the pace of the internet. What I built Obooko is ad-supported like YouTube. No paywall, no subscription, no account required. 4,000+ books across 30+ genres. You can read in the browser,

Article URL: https://www.doanything.com/landing Comments URL: https://news.ycombinator.com/item?id=46906637 Points: 1 # Comments: 0

Article URL: https://www.tryautropic.com/ Comments URL: https://news.ycombinator.com/item?id=46907066 Points: 1 # Comments: 1

Hi HN, I’m the author of The 2025 Entrepreneurship Research Playbook. I am an academic (SOAS University of London) who usually works in public policy. In 2023, I designed a "mini-MBA" for founders and realized a massive inefficiency: my students were dealing with problems that scholars study every day (liability of newness, signaling theory, dynamic capabilities, etc.), but they never saw the research because it was locked behind paywalls or written in dense academic-speak or just too recent to

I got frustrated with existing DJ metadata tools acting as "black boxes"—uploading my audio to cloud servers and giving confident (but often wrong) key detection for minimal/tech-house tracks. So I built Atomic Afterglow. It’s a Python-based analyzer that runs in the browser but processes audio locally (zero retention/no uploads). How it works: Engine: Built on Librosa and Streamlit. Drop Detection: Instead of finding the "quietest" section, it uses a rolling RMS window to find the deepest struc

I built Askie after my child asked ChatGPT about snakes and received a detailed breakdown of the world's most lethal species. Weeks of nightmares followed. The core insight: filtered adult AI isn't the same as AI built for children. A content filter catches bad words but doesn't adjust vocabulary for a 5-year-old vs. a 12-year-old, or know when to redirect a conversation. Askie is voice-first (kids talk, not type), COPPA compliant, and gives parents full conversation transcripts. Image generatio

We've been building an AI agent that debugs production incidents, and we think most AI SRE tools are solving the wrong problem. Everyone's focused on the reasoning — better prompts, better models, RAG over runbooks. But in our experience, the bottleneck isn't the AI's ability to think. It's access to data. At any company past 50 engineers, half the context you need during an incident lives in internal tools with no public docs, custom deploy systems, homegrown dashboards, internal CLIs. No vendo