Only when AI can read my mind. It’s not a matter of being “better” as models are already good enough. It’s a matter of communication. Sometimes you explain your idea using text and model makes mistake. Then you realize that your instructions were ambiguous. The same happens between people. Even if you have spec and give it to another developer sometimes there is this one detail that might be crucial to the solution that you have in mind that wasn’t put into words.
Recently I read that someone created a “language” for the cross-model communication that doesn’t use words and of course saves on tokens. I think we need BCI.
@eualmeidazs There is already a built in query engine similar to Prisma https://t.co/1AUUP0Yn72. Also instead of creating a wrapper it could be just a few utils that you use in the where clause. IMHO that’s a better approach as it’s more composable.
@ekonomat_pl Poczekajcie jak się o tym dowie skarbówka. Jeszcze dowalą im karę jak było z rozdawaniem bochenków chleba przez piekarnię. Ten świat schodzi na psy.
I use paid versions of Grok, ChatGPT and Claude. I don’t do extensive comparisons but it definitely feels like Grok is worse in some aspects. One that is super obvious is image understanding. Grok just makes things up. Another thing is image modification. Recently Grok also had to increase safety for image modifications which sucks. It used to be the least restrictive. Recent example, my daughter wanted to check how she would look like with big muscles - denied because can’t modify child appearance this way? In general for simple answers Grok is fine and produces comparable results especially if it goes about latest events.
@ProconsulPL Dopłaty powinny być tylko w branżach kluczowych dla funkcjonowania Państwa. Niestety jest też tutaj wiele nadużyć. Np. taki człowiek z orkiestry wojskowej idzie na wcześniejszą emeryturę i dostaje ogromne pieniądze.
Yeah, it's annoying that you can't do does funny/unharmful modifications. My daughter wanted me to create image of her all in colors of rainbow using ChatGPT. We started with hair and it worked. Then everything else. And we could do all but skin. WTF? Then realized yeah it's probably because of changing black to white and vice versa so they are super sensitive to those. It's a joke.
I would like to see something that shows why would I have to use Frame for the particular use case as well as client and non-client components. When I was creating my first Remix 3 app, I've started using <Frame> but then realized that they introduce more problems and stopped and everything still worked fine. Also, would like to see some bigger app with a lot of client components and how you organize them without dumping everything into assets and having organizational mess.
@ryanflorence@colmtuite Just to be clear I'm talking about the submit button. I'm not sure about the disabled menu options. I've just assume that it might be the same rule
I've been porting some small website to Remix 3 and I wanted to use your db package but it's lacking a lot of features that Drizzle has. I guess, with time you will implement them but it was a deal breaker for me and I stuck with Drizzle. One is being able to have camelCased field names in code but snake_cased in db. And providing custom field TS types so that I can have branded ID types.
@mjackson Oh I didn't know that there is a diff viewer :D. Just found it in the app. Seems like it should be more apparent in the UI. I will try that then, thanks! :)