The poster gives Claude work tasks and has it implement most of the functionality, though the last 10% still takes the most time.
The poster gives Claude work tasks and has it implement most of the functionality, though the last 10% still takes the most time.
The poster built a simple .NET console app around Claude CLI and GitLab. Every 15 minutes it pulls assigned issues, classifies whether each one is ready, drafts a GitLab reply for unclear tickets, or starts Claude on the repo plus the issue text and image attachments; Claude then works on a new branch, opens a PR, and later applies PR comments, while the poster manually reviews and adjusts drafts before posting or merging.
They say it automated about 80% of the job and cut their hands-on time to roughly 2-3 hours a day for review and testing. On a recent masterdata feature, Claude produced most of a 2-3 hour task in 15-20 minutes, but the poster says it only works because they check every PR and still fix issues like style-guide adherence.
Their practical catch was usage limits: they were on Claude Max, only started new implementations with more than 20% tokens left, and let the loop run overnight.
Here is a work task: [describe the task, requirements, and any relevant code or docs]. Implement as much of it as you can, then list what is left unfinished or uncertain, especially edge cases, testing, and integration points. Finish with a checklist of the remaining steps I should handle myself.
Score every demo request against your ICP, update HubSpot and ping the right rep in Slack within a minute of the form submit.
Every new Zendesk ticket gets a category, priority and sentiment tag plus a draft reply in an internal note before an agent opens it.
Copilot and Claude review each PR for bugs, missing tests and risky changes before a human looks, so reviewers can focus on design.