Lately, I've been researching something that interests me. I still don't know how to make the most of it. It's Amplifier— it isn't simply another tool for coding with AI. It's one of a kind, and I want to share my opinions because they matter.
The Part That Made Sense
What struck me was this: what if we stopped teaching AI what to construct and instead taught it how to think about a problem?
Amplifier calls these "recipes for thinking". In short, you write down the steps an expert would take to think things through.
Instead of saying "write me a blog post," I explain how I really think:
First, look at my past posts to figure out how I write.
Write something new that goes with that style.
Compare that to what I have.
Make sure it sounds like me.
Get feedback and correct any problems.
That's all. Just explain what you want to do, and Amplifier will take care of the hard aspects, including resolving errors, managing state, and async operations. It does the boring task that usually takes me a long time to fix.
What I've Been Trying
There are various example scenarios in the repository, and I've been trying a number of them out. Some of the results really did surprise me: