@Saanvi_dhillon Hardware. Reliability. Someone go make a similarly robust design that is as reliable, see if they can make it cheaper and better, probably not possible. Laptop=Macbook, Desktop =Linux.
@moiiikaaa Yes, developing software isn't about writing code, that's a misnomer, that because until recently developing software needed "coders" they thought "Ah, coders are the only software developers". There are dozens of unique roles outside of "coding" that all count.
@rezoundous Good luck, I "might" go from $200 to $100, but I can afford it, use it a lot (Codex) and we still have the balance of a month left at 20X so, I think I'm sticking with it, then I'll go to $100 and try to build like crazy to use my 62.5k credits.
@vibeonX69 100%? Why would any be written by hand now? That's kind like, "How many emails am I comfortable writing on a computer? Nah, save the special ones for the type writter still ;)" It's the same to me.
@kamilbaranek_ Light because its game changing, its the fastest and it's still amazing. Going to miss it when I don't renew but not getting enough out of it (not enough time to use it at the rate I am, I'll switch to $20).
@dark_coderz Well you're missing that OpenAI announced Cursor won't have its models come November time so, you'll need an OpenAI account if they have models worth using then. But I agree Cursor will probably be how I access Claude, very soon.
@DrewWakefield5 I’m defaulting to $100 from $200, if I ween myself off my other plans then I may stay on $200 another month. My setup will probably be $100 ChatGPT and $20 Cursor, I could do that in perpetuity. Cursor would fill in the heavier weeks if I really needed it, and give Claude models
Hear me out for a sec.
Rather than a $500 plan and ripping people off on the $200 plan, how about a $50 plan instead?
$100 is too much when you're building around work, and $20 is useless without a mid-week reset.
Would a £50 plan work for anyone else?
@jpshrodinger This is exciting if true, will be good to have Gemini back in after taking 2026 off. Today this statement isn’t true but when their pro model launches, competition is good
@xoaanya Data. Compute. Time. Understanding of weights. I think most people on here could get an Open source model, fine tune it, issue is getting mass amounts of training data, storing it, storing the massive models, then assuming you can, how do you train? What compute? It’s quite hard