Peter Steinberger, OpenClaw project steward, details a workflow using GPT-5.5 and Crabbox to scale agent execution up to ten hours
The multi-hour runs significantly increase confidence in production-ready output.
With GPT 5.5, /goal, autoreview and crabbox my prompts moved from ~30-60min to often 4-10h tasks and my confidence that it’s ready is much much higher.
Yielding agents is a skill.
Autoreview: https://github.com/openclaw/agent-skills/blob/main/skills/autoreview/SKILL.md
crabbox: http://crabbox.sh
Peter Steinberger 🦞@steipeteWith GPT 5.5, /goal, autoreview and crabbox my prompts moved from ~30-60min to often 4-10h tasks and my confidence that it’s ready is much much higher. Yielding agents is a skill.
@iruletheworldmo very much depends on the skillset of the person driving the AI.
🍓🍓🍓@iruletheworldmo@steipete do you think we’re heading to a world where this level of work gets done much quicker or longer and longer time horizons but we’re happy because it’s so well done and we’re out of the loop. or both
@steipete do you think we’re heading to a world where this level of work gets done much quicker or longer and longer time horizons but we’re happy because it’s so well done and we’re out of the loop.
or both
Peter Steinberger 🦞@steipeteWith GPT 5.5, /goal, autoreview and crabbox my prompts moved from ~30-60min to often 4-10h tasks and my confidence that it’s ready is much much higher. Yielding agents is a skill.
@steipete but five years from now do you expect ai to be doing a week of work in ten seconds or a year of work but it takes a week to run
or you still think the prompter matters in the limit
Peter Steinberger 🦞@steipete@iruletheworldmo very much depends on the skillset of the person driving the AI.