Skip to content

Comment on Preliminary data from a longitudinal AI impact studyparent

Comments

Planning might end up being more reliable thanks to coding agents: if you want to estimate how long a task would take, just send an agent to do it.

If the agent comes back in a few minutes with a tiny fix, it is probably a small task.

If the agent produces a large, convoluted solution that would need careful review, it is at least a medium task.

And if the agent gets stuck, runs into architectural constraints, etc. then it is definitely a hard task.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.