The challenge of proactive AI agents beyond coding
A user says they still haven't seen a good non-coding product where most actions happen without human prompts by default. Machine-learning training runs are the only use case that makes sense to them so far.
TLDR
In July 2026, a user described proactive background agents outside coding as difficult to build, saying they hadn't seen a shipped product where most actions happen without human prompts by default. Revisiting that observation on September 26, they said they still hadn't seen a good example. Machine-learning training runs were the only use case they found plausible: the agent can simply wait for a run to finish. They also argued that token revenue has limits if it relies on “1% of humans doing the prompting all day.”
Combined views
—
1 Source, first seen ago
