Anyone know of something like this?

How would you have written it differently

I've been preparing for a trip to the Basque Country later this year by learning Euskara (aka Basque). A big part of my learning process is following Euskara-language people on social media, then putting their subtitles into a machine translation service. However, since Euskara grammar is so different from that of English & Spanish, I found myself wanting to know the "how" and "why" of the translation. This visualizer scratches that itch for me.

I'm using a processing pipeline for this app that goes like: 1. Submit the input phrase to https://batua.eus (a Basque-owned & -operated machine translation service) 2. Run both the input and output through Stanford's Stanza NLP Python library 3. Pass Stanza's output to an LLM (Claude) to generate the data structure that drives the visualization.

I'm posting this knowing there are a couple limitations: 1. My API token for batua.eus is fairly limited, so now that I'm posting it to HN I expect that limitation to get hit 2. I'm risking some charges for my Claude API token, so I've had to limit folks to max 10 translations 3. I'll take this down the moment costs get out of control, tho I reckon the limited API token for batua will protect me :)

The monorepo source code can be found here https://github.com/mattdeboard/itzuli-stanza-mcp

It's split out by backend/ and frontend/. The backend has an architecture document I'm proud of. https://github.com/mattdeboard/itzuli-stanza-mcp/blob/main/b...

I've used Claude Code extensively for this project, and am very pleased with the quality I was able to coax out of it. Keeping an attitude of "senior dev mentoring a junior dev" toward Claude, and bringing my own taste/standards to the table.

Comments URL: https://news.ycombinator.com/item?id=47289169

Points: 1

# Comments: 0

But it got me interested in a topic I have been calling "token economization." I'm sure there's a more common term from it but I'm a newb to this tech. Basically, how to optimize the "run rate" for token utilization per request down.

Have you taken a stab at anything along this vein? Like prompt optimization, and so on? Or are you just letting 'er rip and managing costs by reducing request volume? (Now that I've typed this comment out I realize there is so much I don't know about basic stuff with commercial LLM billing and so on.)

[1] https://github.com/mattdeboard/itzuli-stanza-mcp

edit:

I asked Claude to educate me about the concepts I'm nibbling at in this comment. After some back-and-forth about how to fetch this link (??), it spit out a useful answer https://claude.ai/share/0359f6a1-1e4f-4ff9-968a-6677ed3e4d14

I may take a crack at this tool next week.

I know that's been a meme since forever, but my first hand experience supports it to the extreme.