how much value is in the paid version? its hard to understand. my main concern is not knowing how much time ill have to dedicate to get the real value out of it. im running two businesses, and have a variety of big big obligations. and i have deadlines everywhere. but i cant help feeling that ai is the solution to this. but it brings its own dichotomy: will i maybe just spend alot of time, and realise, the reach and speed i was hoping to gain isnt saving me as much time as id hoped because id overstimated it? and maybe i just sink alot more time than i was thinking, actually working in the opposite direction that i was hoping basically. im hoping to get a proper all in one ai brain system that looks over everything in my notes and business data overnight, find issues, contradictions, strengthen a relationship graph, and make suggestions. for both my personal life and my business. but also be able to act if it can. now my issue is, i know most of this is possible. but when im doing it, it gets super messy, and seems like a huge load of time to get right. if someone could just give me 1 answer it will solve my decision. will paying give me the answers or skills to make this, and will it be a non enormous job to get this daily to-do/brain/personal assistant working? i think if its possible to have a brain basically accessing and running individual other automations that have their own project brain. then its worth me to invest. but honestly im not sure whats even possible, and moreso, im not sure if the info is here in these forums on the paid version

so. how are people dealing with this issue, that vscode seems to: Always write to c(lower case) and not C(upper case) file location, & searches for existing in the same fashion. - regularly causing issues with "mcp not found" or other such things, - sending claude into a spiral of context & token waste trying to solve a problem that it never will without you specifically telling it whats wrong. Powershell or Cli however, always use C (with a capital), and alot of the time, claude will automatically run one of these mid solution if i set it to solve something without me. creating two unreconcilable branches. maybe you havent noticed it, because if you have something installed or implemented at ~user for global install, it never has to go through the full file path that uses c or C. meaning global installs are immune to this. it only happens in project specific installs ive noticed Nate doesnt have these issues when hes installing mcp's in his project. it just works directly for him, which makes me wonder what im missing. he is definitely doing something to solve this issue directly. but i don't think hes posted anything about it. i know many people will encounter this problem, so i could really use some help. - its killing all my skills and self healing

anyone have any insights how nate is using perplexity? i have seen a few ai influencers using it, and i can only see it brushed over in any of nates tutorials. maybe ive missed it. perhaps ironically perplexity hasnt helped me find where nate has spoken about it 😅 i see the mcp, but to my understanding the use of sonar is something else. and for the way id be using it, it would be expensive. so instead im currently forced to type everything into perplexity myself, then copy paste the results back to claude, which seems unnescesary is there no other way? are people just using perplexity search which is much cheaper, and using skills with claude to get a similar result? and what about these github repo's that claim to use a workaround for perplexity api where its free. or even some of the github repo's that say they just do what perplexity does but free? any insights?

it seems like these different agents and tools have some differences in how they are holding their mcp, connector or plugin data. what that means is that if you build a self learning/self healing skill with one, and then you use the other agent or tool and continue working, they wont see the correct mcp, connector or plugins available, and will mess everything up! i am certain nate has an obvious solution for this. @Nate Herk i would love to see a video on this. - you are always using vs-code, but claudes connectors and plugins from the desktop app are largely not innately working with vs code or terminal. theres obviously a required workaround for this

has anyone built a skill or hook to automatically reference the latest ai innovators, before doing solving any build plans. i.e. ive seen nate say to claude several times: think of any mcp's or skills that might be useful to use here. however claude is a bit stupid. its training isn't up to date. i need to actively point it to where to look. but obviously its not going to read everything -> assess how everything slots together -> make all these different ideas compete for the best one in that process and even if it did, it would pollute the context like crazy. even using subagents to do alot of that for you would take time. i see two options. and im not in the loop enough to know if people are alraedy doing this or not: 1. make a basic skill that tells it specifically which things or spots to reference - Then create an automation which feeds it - I.E occasionally looking at all the ideas, comparing them and updating a db that feeds the source for the initial skill (could also be used for all domains of knowledge work to great effect) - and more importanly: - - could be built and run by one person, and just accessed by many without any need for everyone to constantly do this "2." create a skill that delegates a bunch of sub-agents to do this every time obviously checking the lates ai innovators methods and successful work/agent brains documentation(if they exist) would just improve your agents abilities to get everything and anything done - assuming you dont stuff everything in context and pollute it