Hacker News
new
past
comments
ask
show
jobs
Show HN: Aharness – Enforce coding-agent workflows as state machines on Codex
(github.com)
4 points
by
alfredvc
9 hours ago
|
2 comments
by
rishabhpoddar
8 hours ago
|
next
[-]
As models get better and better at following instructions, don't you think this will become lesser and lesser useful?
reply
by
alfredvc
8 hours ago
|
parent
|
[-]
The better the models, the fewer guardrails they will need. However, the more you know about the process the more useful the guardrails are. Guardrails may become less useful over time, but you pay for it in uncertainty and likely token usage.
reply
by
tangweigang
5 hours ago
|
prev
|
next
[-]
[flagged]
reply
by
milan_
4 hours ago
|
prev
|
next
[-]
[flagged]
reply
by
codelong888
4 hours ago
|
prev
|
[-]
[dead]
reply