It’s definitely badly communicated and suspicious, I just called out jumping to extreme conclusions based on a suspicion alone. There probably will be people who are gonna review the code and see how much of it is probably LLM generated, and then we will know. I still think that it’s pretty much impossible to vibe code something on that scale, but I haven’t seen their cursorrules either.

Absolutely, I do them daily, what about you?

I’m a pretty big generative AI hater when it comes to art and writing. I don’t think generative AI can make meaningful art because it cannot come up with new concepts. Art is something that AI should be freeing up time in our lives for us to do. But that’s not how it’s shaping up.

However, AI is very helpful for understanding codebases and doing things like autocompletion. This is because code is less expressive than human language and it’s easier for AI to approximate what is necessary.

I’m personally scared of AI (not angry or hateful, actually scared by just how fast it’s advancing) and that definitely clouds my judgement of it and makes nuance difficult.

It’s like a deal with the devil. You see all these amazing benefits but you just know you’re the one being taken advantage of, because, like the devil, AI corporations by definition only think about how you can be of use to them.

You’re not alone. Nuance is just harder to convey, takes more effort to post something nuanced. And so people do it less, myself included. But I think truthfully that many people are not so stuck in one or the other circlejerks. It’s lovely to see people in this thread who are annoyed by both.

Natural language processing makes TTS way more usable for people with reading disabilities. But there are absolutely no good uses of AI.

The best you’re going to get out of it is it shortens the amount of time wasted on tiny adjustment to the UI or something.

This gets into the question of what, if anything, AI “should” be used for.

I’ve heard responses to this go both ways. Some people argue that saving time on repetitive simple tasks is what AI “should” be used for; but other people say that if you can’t even do something as simple and repetitive as a tiny adjustment to the UI, you shouldn’t be in a development job to begin with; or that you’re stealing the work of other programmers who had their code scraped for training data who are not being paid while you are, and that maybe you should be fired and the people who had their code scraped be hired instead.

IDK what the right answer is, I think this is something I will struggle with for ages while the unscrupulous people use AI for everything and anything.

Cursor is just an IDE (integrated development environment), you can set it up to use all sorts of LLMs either directly through Cursor, or with your own API keys for the sources.

This file content just goes into the initial context to help the LLM act how you want.

It will allow you to see if the AI has made any syntax or runtime errors. It does not tell you about any logic errors.

Logic errors are already the most dangerous kind of programming error, and using AI just makes them even harder to find.

Using AI will only help you with syntax (which any good IDE should already be able to do) and finding information faster than a search engine (but leaving out important context). AI is not useful for programming anything that will be made public.

Yeah, you get immediate feedback, vs a scenario where you have to manually check the “facts” it provides in order to ensure it’s not hallucinating. I’ve had Copilot straight up hallucinate functions on me and I knew that they were bullshit instantly.

I iterate with it a ton and feed it back errors it makes, or things like type mismatches. It fixes them instantly and understands the issue almost every single time.

That’s the trick. Iterate often and always give it new instructions if it does something stupid. Basically be as verbose as needed and give it tons of context, desired standards, pitfalls to avoid, whatever. It helps a ton.

I’ve had the greatest success with Claude. The company I work for basically let us all go wild with a few to trial, and Claude has been the best for all of us—even better than GitHub Copilot.

I pay for my own pro plan outside of work and use the VSCode plugin. I’d say read the quickstart guide and experiment with it. Start off with having it do smaller changes and don’t be afraid to be verbose. The more context, the better. Point it to existing files you want to follow the patterns of and model after; give it links to resources for best practices, etc. You can also use it in “plan mode” if you want to see its proposed approach before it starts editing.

I also recommend leaving it so that each change it makes requires your approval (it will do this by default and you can step through everything). That way you always have some control and if it does something dumb, you can stop it at that step and pivot with a different instruction. Alternatively, if you want to see it go ham and carry everything out without approval at each step, you can enable auto-accept.

Once you get into it, start looking into how to craft instruction files. You can have those at your disposal for things like writing tests, language-specific guidelines and practices, etc. That way you can make sure it uses those as a reference so you don’t have to give it the same instructions over and over with every prompt.

If you hate writing tests, I’ve had really good luck letting it handle that. I tend to use it more for the bulk tasks that suck. For things where I want more control, I work with it on a piecemeal basis in my project.

thats a good point. fascists are always sus in many ways.

the ceo of the company

Not an issue, per se. In order to keep the team small they built most the app in a single codebase. It’s mostly web code, and the apps are wrappers for it. So it keeps it unified between all clients but it definitely feels like a web wrapper, so it can feel a bit slow or clunky.

@sunzu2 it feels janky as hell, it’s missing advanced features (someone in the other thread asked about Sieve filters), and it doesn’t support non-Tuta clients. their development cycle is so slow I can’t count on any of these features cropping up anytime soon.

with those criticisms in mind, Tuta’s still approximately the only credible choice remaining for threat models where end-to-end encryption is important. we desperately need better fully open source options for this.

I have tried Tuta out, it’s fine from my very limited use, but kinda locked in in ways I don’t really care to pay for. Last time I saw it brought up some other folks were recommending mailbox.org. I don’t know about it too much, but might be worth looking into as well.

I also use bitwarden (work), and it only does password management. It doesn’t do email and alias generation.

Yeah it’s not just for privacy, hence the port forwarding requirement.

AFAIK nothing has shown issues with the privacy of either email or VPN? At least not something that wasn’t caused by blatant idiot user error like the guy with his apple email as recovery email.