We will need to hijack dopamine for Education the same way that Social Media has done if we're going to keep up with the machines. Not everyone will like what this means.
People have this purist notion that kids should just be self-motivated to do well. For about 5% of them, that's true. But what about every other kid?
Motivation is 90% of learning. Our guides figure out what motivates each kid and use it to get them to put in the work. Over time that builds a confidence anchor: the one win where a kid did something hard and hit their goal.
Confidence in yourself might be the most valuable thing a kid can build. It changes your entire academic path. And yes, sometimes it starts with an Amazon cart.
@midwitologist@ben_m_somers Content for education is a virtually unlimited resource. The orchestration layer has been what fails most kids (either too slow or too fast, or not enough remedial if they missed something).
@kimmonismus@ZikuD_s It's a function of how the models inherently work, since day1. It's actually *why* the models work, not a secondarily discovered byproduct thereof.
You're expected to predict the rules that will apply and coordinate around them. If you can't create a mental map of how opposing counsel is going to behave in an essentially unethical fashion to exploit rules to their advantage, and develop systems and tactics around that, then you should recommend someone who will do that to your clients.
Because our kids are getting their academics done in a shorter period of time, they have much more time in the day when they are connecting with each other, socializing, and working on projects together.
@henrytdowling@justindross Easily falsifiable. Lots of parents who care about and are involved DEEPLY in their child's education that don't obtain results close or similar to Alpha School.
guy selling post trained oss models: if you donโt own your models you donโt own anything
guy selling harness on top of models: if you rely on one model you are cooked
guy selling tokens: instead of manually prompting your models just put them in a loop where they prompt themselves
Imagine being some teenager in 1962 and being told that you would need to be concerned about how much electricity you use in 2026. You would have scoffed at and ridiculed the idea, justifiably. How has this happened?
@marriedmn This is absurd advice. Be willing to fight a nuclear war for your children to protect them from the type of person who would do this to them.
In life, you must choose your regrets.
โYou'll regret it if you get married. You'll regret it if you don't get married. You'll regret it if you have kids, and you'll regret it if you don't.
Kierkegaard said this 200 years ago as follows:
โWhatever you choose, you'll regret it. Because the problem isn't in your choices; it's in romanticizing a life you haven't lived.
A person always finds an untraveled path alluring and mysterious.
That's why the issue isn't making the right choice.
It's choosing and deciding which regret you'll live with.โ
What have you decided?โ โ Salih Guney
@housecor@themikebwebb The solution IMHO is to have a set of AWS secrets that contain the values of the Env Vars, and to retrieve those via AWS Secrets both via GitHub Actions and locally.
Eliminates a lot of that.
Expert Mode is running all of it in Docker in both environments.