Post B7PHvoZXoMWA7Wnm0e by crunklord420@kiwifarms.cc
 (DIR) More posts by crunklord420@kiwifarms.cc
 (DIR) Post #B7PHvoZXoMWA7Wnm0e by crunklord420@kiwifarms.cc
       0 likes, 0 repeats
       
       Seeing the Claude Max plans really made balk at the idea of even trying Pro. I kept hearing how little tokens you get on Pro.You can get a ton of shit done on a Pro plan if you're not just burning 100k tokens vague-posting your way to a MVP. I don't feel that limited by the quota.
       
 (DIR) Post #B7SqcMlAar3UB81lsO by Saxophone3784@kiwifarms.cc
       0 likes, 0 repeats
       
       @crunklord420 The issue with Pro is only if you deal with software that has a lot of files. for example, I use Max to map out an abandoned large open source project to rewrite it in C. That is nearly impossible on Pro.
       
 (DIR) Post #B7SqcN0lerX2xVKEBk by Dirac@kiwifarms.cc
       1 likes, 0 repeats
       
       @Saxophone3784 @crunklord420 Blanket dropping "re-write this entire codebase k thnx plx" is going to hurt the poor AI, structured instructions and isolated segregated blocks of code can unironically save 10x because you never get it in those circular reasoning loops where it goes "oh God part A gives input to part B and then oh Jesus part C changes state from part B and then the user is prompted from part A and then...". It has ADHD, you have to keep that shit on lock.
       
 (DIR) Post #B7SqjXdmXMYZx4zeds by crunklord420@kiwifarms.cc
       0 likes, 0 repeats
       
       @Dirac @Saxophone3784 Very true. I always try to specify a deliverable. I've seen LLMs waste reasoning going in circles trying to deduce implied instructions. You can short circuit all of that by giving a good prompt.