# Developer urges AI coders to improve existing open-source tools

A software consultant is calling on developers using generative AI to direct more of their effort toward established open-source software, arguing that practical improvements to widely used projects could deliver more value than another wave of small standalone applications.

Jordan Krueger made the case in an essay published on November 17, focusing on the operational barriers that still prevent small organizations from adopting free and open-source tools. AI-assisted coding has lowered the effort required to build prototypes, he wrote, but much of the visible output has taken the form of new habit trackers, task managers and note-taking applications.

Krueger’s proposal is to apply that same capacity to less conspicuous maintenance work: reducing memory use, improving installers, expanding test coverage, clarifying error messages and completing documentation. Those changes may not produce a new product, but they can improve software already used by large communities.

The argument draws on his experience running self-hosted services and advising progressive nonprofit organizations. He described maintaining a Matrix deployment serving thousands of users, as well as operating software through hosting tools and low-cost infrastructure. In that setting, the core applications are often capable enough, while setup, upgrades and troubleshooting remain difficult.

That complexity has consequences for organizations interested in controlling their data or reducing dependence on large technology vendors. A small nonprofit may prefer an open-source option but lack the staff required to manage it. When installation assumes familiarity with several supporting systems, upgrades fail unpredictably or an application consumes more resources than a modest server can provide, a paid hosted service becomes the easier choice.

Krueger also emphasized that maintainers have limited attention. He cited the burden created when AI tools generate low-quality security submissions, while noting that even legitimate reports can overwhelm a project if their volume rises faster than maintainers can review them. His recommended workflow therefore keeps people responsible for checking AI-generated changes and encourages contributors to coordinate with existing project teams.

The essay presents maintenance as a potentially strong use of coding assistants because many valuable tasks require persistence and systematic verification rather than a novel product concept. Profiling slow code, testing installation paths and writing regression tests can make later changes safer, provided a contributor understands the project and reviews the resulting code.

Krueger’s practical challenge is narrow: choose an open-source tool already in use, identify a recurring problem in installing or operating it, and spend development time fixing that issue with its maintainers. The proposal reframes AI-assisted coding as infrastructure work whose benefits can be shared, modified and redistributed rather than confined to another isolated application.