Prompt engineering crash course — local models, system prompts, context engineering

You got a minute. I'll teach you prompt engineering first up You're gonna want local AI on your computer. It's more environmentally friendly and also open-source and not monitored by Palantir You can run models made by countries that have health care and stuff and run models on this The 16 gigabyte orange pie want the 16 gig it costs 140, but if you're gonna use Amazon I have one of these I guess I would they showed me this when I screenshot it I don't it will run a 7 billion parameter language model and maybe a 14 billion parameter language model quantized down to four bits It's this one with the four bits And you could talk to it like this and it runs locally You can set a system prompt over here tell it not to apologize or be sycophantic Tell it to cite its sources you click over here under program You can add tools to it you can let it use the terminal or you can let it look things up Recommend you take a class on critical thinking but we discuss it in my class because it is critical to working with AI that you know How to ask for actionable and specific things? It is so important that it is day one of the class. There are a lot of rhetorical frameworks You can get a PhD in this you will cover some Once you get a workflow going something that works for you you can save it and you can also make that thing Available to other people use stuff like Langrath or the open AI agents SDK There are like 50 zillion of them the guy who wrote 12-factor agent as they're bad And I…