How to use Claude to upgrade a stalled codebase
Claude completed overdue upgrades on a branch while the author checked in occasionally. The team then performed several engineer days of manual testing and deployed.
news.ycombinator.com
Prompt to try
Editorial adaptation, not the source author’s exact prompt. Replace bracketed inputs and review tool permissions first.
On [upgrade branch], bring [codebase] up to date with [target library and framework versions]. Work through the upgrades while I provide periodic direction. Document changes for our manual testing before deployment.
Reach people exploring AI agents
See placements →What happened
The upgrade work finished in a couple of days and was deployed after manual testing; resource and cost improvements were not yet verified.
Reported by dkarl on Hacker NewsHow the agent did it
- Create a branch for the overdue codebase upgrades.
- Give Claude the upgrade task and check in with further instructions.
- Dedicate several engineer days to extra manual testing.
Where a person stepped in
The author checked in periodically; the team performed several engineer days of extra manual testing before deployment.
A person set the goal and reviewed the result.
The author described a work codebase that had received minor library upgrades but no major upgrades for several years. It had been recognized as significant technical debt for almost two years. The team believed bringing it up to date would require an engineer's dedicated attention for a month or more. They also suspected framework upgrades might improve performance enough to reduce operating costs.
The author created a branch and gave Claude the upgrade work. Claude completed it in a couple of days while the author mostly worked on other things. The author checked in occasionally to provide further instructions, and explicitly noted that this limited attention might explain the elapsed time. They did not know how quickly it would have finished with continuous attention.
After Claude's work, the team spent several engineer days on extra manual testing. The result was deployed. Performance savings were still a hypothesis: the next planned experiment was to give the application fewer resources and see whether the improvement held up in practice. The codebase was approximately 200,000 lines.
Who it's for: Engineering teams with overdue dependency or framework upgrades and capacity to manually test the resulting changes.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · dkarl
Frequently asked questions
Was the upgrade reviewed before deployment?
The team dedicated several engineer days to extra manual testing after Claude completed the upgrade work, then deployed it.
Were operating-cost savings demonstrated?
No. The author said they were ready to experiment with fewer resources to see whether the suspected performance improvement held up in practice.
More details
Remaining steps
- Deploy the upgraded codebase after testing.
About this account
Claude carried out upgrades while the author worked mainly on other tasks and supplied occasional further instructions.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=47615339. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: dkarl on Hacker News
Source published 2 Apr 2026
Source checked 14 Oct 2026
Tools and topics
Claude · Git branch
Software development
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.