Wartime's avatar
Wartime
npub15wzu...skp8
Wartime's avatar
wartime 1 month ago
Wartime's avatar
wartime 1 month ago
What I would like to see more Nostr apps doing is hiding the fact that they are Nostr apps. Just obscure it away so normal people think it's a normal app, and not a power app. And let the features and usability make them a fan without them realizing it View quoted note โ†’
Wartime's avatar
wartime 1 month ago
I've been setting up mine to automatically log conversations. I've been feeding it all of my notes and business ideas. Everything gets automatically cataloged into obsidian. That obsidian vault is included in the QMD database for vector memory search. My agents have excellent short and long-term memory with minimal token usage. To set it up right, you really need to spend quite a while setting it up - or buy the files from somewhere. I've been setting up going on 3 weeks. It has already made me more productive. Soon my calendar will be synchronized. Client leads with automatically be filtered and surfaced front and center to me. The more you use something like this, the more you will start to understand how you can compound little abilities on top of each other. One cron job can trigger a cascade of events and work. I have used simplistic bots for many years. I used to have one that would pay my child support for me. That ran with no issues for about 5 years before they updated the site and broke my bot. I used to use another one for automatically populating descriptions and titles for an online store. At that time, I was replacing an employee with a simple bot. These things have been providing real value for people, for longer than people assume. Again it all comes down to how you use it and what you're doing with it. 6 to 12 months from now the landscape will be much different. Right now it's very hard to set up correctly and easy to burn money if you don't know what you're doing. But soon a lot of these things will be ironed out and much more normal people will be able to get higher value from it. On top of that, every time the models get better or cheaper, that means you can also upgrade your system to be smarter or cheaper. View quoted note โ†’
Wartime's avatar
wartime 1 month ago
I don't know who's spending $1,000 but you doing it wrong. I spent $15 in the past 3 weeks with more usage than I've ever had. It all depends on the model. But If they are also using it to do scraping and analyzing, it's not exactly simple work. And if you compare that to a human being doing the same work - that $1,000 is probably much cheaper than a human spending the time manually scraping and analyzing. I think I saw a graph recently saying that Claude opus outputs - in one prompt - between 5 and 15 hours of human work. Obviously that's not on every prompt, but that's pretty substantial even if it's expensive at a 30 cent API call, then maybe it's an extremely good deal. It's all very relative to what you're using it for and what your accomplishing in the end. Some of us have been playing with AI models for years already so we already know how to prompt better and we get better results. You've got to keep playing with different models - talk to them in different ways - and be explicit, try to remove ambiguity. And if you're having problems writing prompts - ask one AI to write a prompt for you to give to the other.. and that will often increase your chance of a one shot successful prompt View quoted note โ†’
โ†‘