levhawk

Sure, I agree. I was more about encouraging to not be afraid of something hard.

Today's narrative pushes people to try vibe code as much ideas as possible, even in parallel, but I don't think it's a fruitful approach. One should not be afraid of doing something non-typical (hard) if they belive in the idea. And if you believe, dedicate some quality time for it. If you don't believe - why bother even with prototypes?

In the current era of urge to make things faster/more, that article felt refreshing. Good things take effort and time to create, and it is not something bad, it's just the way the things become good.

On token consumption and efficiency... AI-champion guy in my prev company made a metric, like how many tokens are spend per line of generated code, and even put a leaderboard based on that metric, praising guys with the cheapest LOC.

For me that's insanity for so many reasons...

My position was more about access to information, that business side for me is secondary, as unavoidable evil. If kagi will find a ru-index with decent quality, I'll be more than happier. Right now yandex has the best index, but given the decay of ru tech sector now, especially now when it's oligarch-managed, that might not be for long.

Yes, Yandex was effectively seized by Ru govt "friends" and turned into propaganda tool. And that's very unfortunate. But I'm not talking about getting the propaganda, only about search index. About money - don't think this income is even remotely comparable with oil/gas incomes, which EU passes to RU.

Please don't start political debate. I do not like censorship of any kind, hence my initial response. I want to have available information in full.

Ru-speaking audience is ~2 times bigger than Russia, why would they cut this? Only because of SJWs like you? I'd much prefer one search engine that searches well on two languages, instead of using different engines for each language. Ru-net has huge amount of usefulness in it, cutting it out is like cutting a finger. Fun fact, before the RU-UKR war, most Ukranians contributed to the Runet, so that would cut their heritage too.

After your post I tried to find dotnet groups in facebook, but all I found is spam and almost no humans, which again reminded me about dead internet theory. Could you please recommend some groups?

That looks crazy coming from statically-typed languages. So many hops, efforts, custom structures just to verify types. Come on, modern C# can do all of that out-of-the-box. You define a record/class for DTO and System.Text.Json will either convert it successfully or throw you an exception that will say exactly what the problem was and at what character/field. Combined with much more advanced IntelliSense, development comfort is so much better. But of course, whatever works for you.

So, there's a hope, cool

Just for the reference, what I have to face with: "Sizinde Sular kokuyormu çok kötü koku var baneda mutfakta"

NLLB gives: "You smell like water. It smells bad in the kitchen."

The right translation is "Do your waters also smell? There is a very bad odor in my kitchen too."

"Bu bina elektronik bir sistem var dış kapı -otomatik- uydu -asansör elektrik yoksa hiç biri çalışmaz yanlış sigorta kapatılıp açılırsa böyle sıkıntı olur"

NLLB: "This building has an electronic system. An external door. An automatic satellite. An electric elevator. If none of them are working, it's a problem."

Should look like "This building has an electronic system with an automatic exterior door, satellite, and elevator. If there is no electricity, none of them work. If the wrong fuse is turned off and on, such problems occur."

They have regular punctuation marks like dots, commas, questions, and exclamations, but, for some reason, they don't use these in chats. I don't know why, but it hardens translations a lot.

I bookmarked/starred the repo to track the updates :) Just to clarify - no pressure, obligations, expectations, etc. Just my feedback with couple of samples, hope that helps.

I like the idea, it's awesome! Too bad that NLLB, despite being advertised as specialized translation LLM, doesn't work with Turkish - it generates rubbish like Google Translate, which doesn't make much sense. I live in TR, but don't know Turkish :( I tried all major translators, for now the only option which performs well is ChatGpt. I created a custom GPT there, instructed it to be my translator, and that works so well, even recognizes poorly written messages without punctuation. Anyway, this is just my observations. Best of luck with the app!