Ancient Roman Board Game

Play Ludus Coriovalli free online — a 1,800-year-old Roman strategy board game lost to history, now reconstructed by AI from a mysterious carved stone. 4 Hounds vs 2 Hares. Published in Antiquity (2025). Play against AI or friends.

Apple's new SpeechAnalyzer API, benchmarked against Whisper and its predecessor

We measured Apple's new SpeechAnalyzer against SFSpeechRecognizer and Whisper on 5,559 LibriSpeech utterances. Apple never published accuracy numbers. Here they are, with raw transcripts you can rescore.

Telegram's t.me domain has been suspended

Whois Lookup for t.me

SalesPatriot (YC W25) Is Hiring Full Stack Engineers (SF)

Experienced full stack developer responsible for building the application's core infrastructure and features.

The art and engineering of Sega CD Silpheed

TFTP Honey Pot Results

My TFTP honey pot has been running for over a month, continuously on my $5 a month VPS, and intermittently on my Dell R530 home server. It’s time to see what surprises it has captured.

The Origins of Heikki's Garden of Flowers

An archive of ~2500 pictorial typography and letterpress works spanning four centuries, images composed entirely of type ornaments, characters, and rule. The history of ASCII art before computers.

Linux on the Sega 32X. Who needs hardware synchronization primitives anyway?

How wanting more is not always better

Climate.gov was destroyed. Open data saved it

"After losing their jobs at NOAA, Rebecca Lindsey, her sister and another colleague teamed up to rebuild a pivotal resource the Trump administration took offline."

Show HN: Nobie – an Excel-compatible runtime for agents and humans

A voxel Tokyo in real Japan time – ride the Yamanote line and study Japanese

Ride the Yamanote line through a voxel Tokyo synced to Japan's real clock, weather, and seasons. A lofi bed plays while N5 sentences are read aloud and drift by as subtitles — an ambient Japanese study room you just press play into.

The 4-Bitter Lesson: Balancing Stability and Performance in NVFP4 RL

We have developed and shared a low-precision RL recipe preserving higher-precision training dynamics. In this recipe, we needed to address instability from the forward pass due to policy quantization errors, from the backward pass due to gradient mismatches, and at the intersection of both due to a small set of particularly sensitive weights. We explain how we addressed each and validated the final recipe.

Thunderbird Desktop settings research: what we learned from your feedback

Quick overview of the insights gathered from interviewing 10 users about how they manage preferences and configurations in Thunderbird desktop.

Building and Shipping Mac and iOS Apps Without Ever Opening Xcode

describe

Show HN: YouTube Guitar Tab Parser

CLI that turns a YouTube guitar-lesson video into a PDF of the guitar tab. - marcelpanse/youtube-guitar-tab-parser

Linux 0.11 rewritten in idiomatic Rust, boots in QEMU

Linux 0.11 rewritten in idiomatic Rust: kernel, std-style user library, and 60+ coreutils — boots on i386 in QEMU. - Poseidon-fan/linux-0.11-rs

Show HN: Jacquard, a programming language for AI-written, human-reviewed code

Jacquard is a small programming language designed for a regime in which most code is written by machine-learning models and reviewed by people. - jbwinters/jacquard-lang

Show HN: I implemented a neural network in SQL

An experiment to query Xarray datasets with SQL. Contribute to xqlsystems/xarray-sql development by creating an account on GitHub.

The real prices of frontier models. Tokens * Price, right?

If you build with an AI coding agent, your workload is basically TypeScript - and that's exactly where the model tokenizer gap is widest. The same TypeScript file becomes 1,178 tokens on Claude's newest tokenizer versus 681 on GPT-5.x's o200k: 1.73x more, before a cent of price difference. We measured every frontier tokenizer to show why $/Mtok is not a comparable price.

Samsung will delete your health data if you don't let them use it to train AI

Samsung has started showing Samsung Health users a controversial notice requiring them to consent to their data being used for AI training if they want to keep their data from being deleted.,Samsung has started showing Samsung Health users a controversial notice requiring them to consent to their data being used for AI training if they want to keep their data from being deleted.