4 comments

[ 0.68 ms ] story [ 19.9 ms ] thread
I had a eureka moment when I realized that harness engineering is duplicating all the work already put into decades of programming language research.

So I forked a Go harness (Crush) and integrated a Go interpreter (the Gno VM) so that it thinks, plans, remembers in Go/Gno objects.

I tend to agree, though I would put it more as we've pushed deterministic stuff onto the model tasks rather than writing scripts like always.

I have some ideas around CUE for this, and you have an interesting idea here I want to think more about.

Thanks for the share

The next video will show an agent swarm around a tree of subtasks. No databases needed for coordination.
I'm using this phrase "hands free coding vs eyes free coding"

my goal for a harness and AX may be different (I'm in the hands free camp and see serious issues with the eyes free camp)

I dont have any issues with coordination because my agents and subagents are a tree. My idea is to define workflows, agents with determinism around them, and have them sit in the same place as agent definitions in OpenCode (as an example)