rbren
Joined 2 years ago
Joined 2 years ago
We need to define agents in code, and drive them through semi-deterministic workflows. Kick subtasks off to agents where appropriate, but do things like gather context and deal with agent output deterministically.
This is a massive boost in accuracy, cost efficiency, AND speed. Stop using tokens to do the deterministic parts of the task!
A logical reframing is not equivalent though! We know everyone else gets the same frame, and most of the problem is predicting what other people will do when presented with this particular two-button frame.
* many people (at least toddlers, people with dementia) are going to press blue roughly by accident. See the lizardman constant
* other people will not want to be responsible for any deaths and will press blue out of a sense of moral imperative
* many other people are going to take this into account and vote blue out of hopes we can save everyone
You should vote blue.
If you want to play with the basic building blocks of coding agents, check out https://github.com/OpenHands/software-agent-sdk
Here are the prompts I use for my AI environment, though it's changed a bunch since the last snapshot