Urgent.News

What's breaking now, across thousands of outlets.

Tech

I asked Gemma for Japanese homework. It wrote about eating a friend.

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle What I Built JaExC (Japanese Exercise Creator) is a Ruby gem that writes kanji-reading exercises with a self-hosted open-weight model, and then refuses to trust it. My partner teaches Japanese. She asked for example sentences using specific kanji, which inspired a larger teaching project. That project predates…

Gemma, a Japanese exercise creator, produced a writing about eating a friend. This Ruby gem employs an open-weight model to generate kanji-reading exercises, then denies trusting its own output. The project was inspired by a request from the creator's partner, who needed example sentences using specific kanji characters. JaExC is a standalone gem that potential customers—Japanese teachers—could use to reduce the time spent preparing exercises for students with different kanji knowledge.

The tool generates sentences with kana readings and English and Polish translations based on the target kanji and allowed kanji lists. It does not validate the naturalness of the Japanese or the accuracy of the translations; a teacher must still review the exercises before use.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in Tech

I shipped a redaction library, then found it leaking a private key

I write a small zero-dependency library, llm-stream-guardrails , that strips secrets and PII out of an LLM's output stream before it reaches a screen or a log. Last week it was leaking a private key.

  • Library llm-stream-guardrails leaked private key due to bug
  • Bug allowed cleartext streaming of PEM private key
  • Fixed in version 0.7.9 with adversarial testing

YouTube transcripts for RAG: fetch, chunk and embed captions in 30 lines of Python

Disclosure: I built the Apify Actor used in this tutorial. This post was written with AI assistance. A talk, a lecture series or a product channel is a knowledge base that happens to be stored as…

  • Install Python 3.11+, Apify-client, sentence-transformers, and numpy packages
  • Fetch YouTube transcripts via Apify API token, segment into 60-second chunks
  • Embed chunks with SentenceTransformer for semantic search and retrieval

Which mule accounts do you freeze? A checked minimum cut on a 4.49-million-transfer FalkorDB graph, and a max-flow bug I hit

After a scam alert lands on a mule account, "is this account bad" is already answered. The useful question comes next: the money is moving, so which few accounts do you freeze so it loses its routes…

  • Chokepoint tool freezes mule accounts to stop money laundering
  • FalkorDB graph had maxFlow algorithm bug affecting 12-17% of networks
  • Issue resolved by adjusting capacities and adding comparison check

More from Sunday 18 October →