We don't need more apps. We need better open source
- Source
- jordankrueger.com
- Author
- pandemicsoul
- Date

Context
Jordan Krueger, who self-hosts his own services and handles technology for progressive nonprofits, argues that the energy now going into weekend apps built with AI would help far more people if it went into the open source software they already depend on (jordankrueger.com). His complaint is not about what that software can do. He says free and open source tools already cover most of what he and his clients need, but running them is the hard part: install guides assume knowledge of five other tools, upgrades break things, and some apps use 4GB of memory to serve a handful of people. His nonprofit clients would like to leave big tech platforms for privacy, cost and control of their data, but the open source option usually needs a part-time system administrator they cannot afford, so they pay for the corporate product.
He proposes four kinds of work: making a tool faster, making it install with one command and sensible defaults, making it easy to run every day through a simple admin screen, safe upgrades, built-in backups and error messages that say what to do next, and making it light enough for a $5 server, an old laptop or a Raspberry Pi. His model for speed is uv, a Python package installer that, he notes, its makers benchmarked at 8 to 10 times faster than pip. The reasoning is about reach. A new to-do app helps the few people who find it, while making a tool that millions use 20% lighter helps all of them. He thinks this patient work suits AI coding tools well, such as profiling slow code, writing tests, testing installers and filling in documentation, provided a person checks every line.
Krueger attaches conditions because maintainers are already stretched. Many are unpaid volunteers, and he recalls the 2024 xz backdoor, where an attacker spent years earning the trust of the exhausted solo maintainer of a compression tool found in most Linux systems. He also cites the curl project, which ended its bug bounty in January 2026 after a flood of fake AI-generated security reports; by his account, curl's lead developer later said the fakes had mostly stopped and the problem had become sheer volume, even of good reports. So he asks contributors to use the software first, talk to maintainers before writing code, bring measurements, keep changes small and tested, disclose their use of AI and stay to fix what breaks.
This is one practitioner's argument from his own clients and servers. He offers no figures on how much AI-assisted effort goes to new apps compared with existing projects, and the 20% example is an illustration, not a measured result.
Checking sign-in…
Loading comments…


