@Tutanota@mastodon.social I second @seanm@infosec.exchange's reply to this post. I'll be leaving your platform if LLM code starts being widely integrated into Tuta apps; it's proven itself time and time again to be not only hugely buggy, bloated and inefficient but an *enormous* security risk and it's absolutely not something I want in my apps that aren't too big, run reasonably well and do their job as advertised. *Actual* (not LLM-generated or "assisted") code, code that's actually been read through as it's written, is always going to be preferable to hastily-generated code that is genuinely not possible to check through and correct by hand (because if you're actually putting in any effort to check it it will need extensive fixing) adequately without the time savings being null and void. I am BEGGING Tuta to do the sensible thing here and take a firm position against the use of generative AI in its software. Tuta is such an agreeable company and it would be a damn shame to see them screw themselves over like this. Even if it's only used for "repetitive processes" (it always starts there of course), I thought we all agreed that a thousand lines of potentially dangerous bloat and unnecessary "features" was the LESS efficient option regardless of how fast it was written compared to a hundred lines that do as advertised and do it really well, as Tuta does at the moment. @jab01701mid@mastodon.social @pyratebeard@harbour.cafe