MichaelHoste
Joined 14 years ago
Joined 14 years ago
Yes, we're using ChatGPT 3.5 for now and we will certainly switch to GPT-4o mini in the near future. It already works with ChatGPT 4 but we noticed that the cost increase was not worth the minor improvement of quality.
I tried your example with ChatGPT 4 and it's somewhat better but not perfect: "I've even come to hate my brother and younger brother."
We hope that the quality will improve with the cheapest (and also fastest) models available.
The legacy software could explain it. But the difference of processes between my main bank (random temporary PIN obtained with 2-factor authentication: postal letter with code to send back using smartphone) and my credit card company (bank card mailed to my address with current PIN) was weirding me out.
But hashing them with the customer ID, date of birth, or something else, and using bcrypt to slow bruteforce down could already be an improvement.
Here is the press release from Blender: https://www.blender.org/press/ubisoft-joins-blender-developm...
If you use React, one of our users also created a plugin for React-Intl: https://github.com/deecewan/translation-io
We think that every framework must have a tailored solution to handle translations best, so we move forward in small steps to be sure our solution fits the pain (of maintaining translations up-to-date as the app evolves).
Even if we compared GetText to YAML key/values (Ruby on Rails), the same reasoning applies also for PHP key/values of Laravel.