News for hackers. Stuff that matters.

loosetypewriter

Show HN: Sigmajp – Integrated Japanese Reader

https://sigmajp.works/

I always recommend yomitan and Anki to other intermediate Japanese learners, and if they try it they love it, but they only try it if I set it up for them.I combined an epub reader with a Japanese text parser and dictionary so that tap-to-lookup a word works with no setup. Then I added flash card creation and spaced repetition reviews. And a built-in 'explain' button that asks an LLM to...

( Read More... | no comments | on HN )
Re:Playing "Taboo" with AI Watermarking on Tuesday September 08, 2026 @06:15PM (#49617937)
Made produced and created, wrote? imply agency... Anyways my point is about replacing widely used jargon with made up jargon (ok new jargon) that makes you squint and obfuscates the connection to industry standard.

That glossary also reads like this:

Why it matters Without a fabric, integration complexity grows quadratically with each new system. A capability fabric linearizes this growth by providing shared infrastructure for discovery, governance, and composition – turning integration from a liability into an organizational asset.

Anyways, I'm pretty sure the only reason you'd do this (watermarking) is to be able separate your own outputs from the internet. And if you can make it a law so all the other AI companies collude even better.

Playing "Taboo" with AI Watermarking

Today ChatGPT emitted a design doc that described our service to service messaging architecture as a "capability fabric". Notice the double noun, as if the a single 2-part word was thesaurused one word at a time. Unnoticeable they said.

( Read More... | 2 comments | on HN )
Re:Why Copilot Is Making Programmers Worse at Programming on Wednesday September 11, 2024 @03:58PM (#41514838)
If you are an experienced coder, then you know which tasks are useless chores. It's a different issue for newdevs who will be learning to code with tool assistance.
Added reverse cards (only applies to new cards, no backfilling). There's a toggle now in the user settings page (inside the top bar user icon) to enable it. It shows the first sentence of the english text as the front of the card.
My quick fix is to set an input length limit on the chat text. Testing out your input, it seems like around 70 is the limit for getting decent output from the LLM.
Thanks, your feedback is worth way more than $10! I have been typing in one line or a couple sentences at a time. Your use case of inputting a whole email sounds very useful though.

I'm wondering if that would require a different UI or if... Say it just broke up the email for you into separate sentences and input them separately. Then you could click through the sentences one at a time in the existing chat carousel, would that work?

Reversing the flashcards wouldn't be much work, but I've heard that's less useful for learning, since for the English text there's be many ways to say it in Japanese. Happy to add features people will use though.

Show HN: LLM grammar explainer and flash card maker for Japanese

https://goginko.com/

Goal was to make it easy to jump from textbook learning into immersion. I wanted to play video games in Japanese, and started by using LLMs and a notebook, but figured it would be easier to streamline it all in an app.- You input a sentence in Japanese (easier on mobile where you can handwrite kanji you don't know) - App translates the sentence, explains the grammar, and gives you a list of...

( Read More... | 8 comments | 2 points | on HN )
Re:How to See Bright, Vivid Images in Your Mind’s Eye on Sunday October 01, 2023 @12:18AM (#37722490)
Can people who literally "see" clear mental images look at a blank piece of paper, imagine an image, and then just trace it out? I have never seen anyone do this, but based on the way mental imagery is described online by people who claim other people are aphantasiac, I don't understand why it's not a very common ability.