Krueger highlights the need for AI‑assisted improvements to open‑source tools that nonprofits rely on.
*Every weekend a dozen AI‑built apps appear, but the real bottleneck is fragile open‑source software. Jordan Krueger argues that AI‑assisted coding should target faster installs, lighter footprints and reliable ops.*
Every weekend a dozen developers push a new AI‑generated habit tracker, to‑do list, or note‑taking app. The flood shows creativity, but the underlying open‑source stack remains brittle. Jordan Krueger, a consultant for progressive nonprofits, sees the gap: free software does most of what his clients need, yet it rarely runs smoothly on cheap hardware, installs in one step, or stays up without midnight forum threads. He runs a Matrix server for thousands, hosts services on Coolify and Hetzner, and stores paperwork in Paperless‑ngx on a Mac Mini that constantly runs out of disk space. The software works; the operations do not. Nonprofits want to ditch Big Tech for privacy and cost, but the open‑source route demands a part‑time sysadmin they cannot afford, pushing them back to corporate tools.
Krueger runs a Matrix server for thousands, hosts apps on Coolify and Hetzner, and keeps paperwork in Paperless‑ngx. The software works, but installation assumes mastery of five other tools. Upgrades break things. A typical app eats 4GB of RAM to serve a handful of users. Error messages give no guidance. Nonprofits lack a part‑time sysadmin, so they pay for corporate alternatives despite preferring privacy and data ownership.
The Rust‑based tool uv cuts a Python developer’s package install time by 8 to 10 times in its makers’ benchmarks, and up to 115 times with a warm cache. Krueger suggests profiling the slowest part of a widely used tool, measuring it, then fixing it. A single‑command installer with sensible defaults can turn a multi‑step setup into a one‑click launch on a cheap $5 server, a laptop or a Raspberry Pi.
If a tool runs on a $5 server, more organizations can afford it. Lighter code reduces energy use and hosting costs. A new to‑do app helps a few discoverers; a tool that is 20% lighter benefits millions daily. The impact multiplies when volunteers maintain the code, because fewer resources are needed for upgrades and backups.
The curl project ended its bug bounty in January 2026 after a flood of fake AI‑generated reports. By April, most reports were AI‑written and high quality, but volume buried genuine work. Krueger advises using the tool first, identifying real pain points, and keeping pull requests small—e.g., a “40 seconds to 4” improvement with clear metrics beats vague claims of “faster.” Transparency about AI use and staying to fix breakages can keep maintainers from being overwhelmed.
He notes that a dozen AI apps appear each weekend, but the real need is better open‑source tools that run on cheap servers and install in one step.
Uv is reported to be 8 to 10 times faster than pip in benchmarks, and up to 115 times faster with a warm cache.
The curl project ended its bug bounty in January 2026 after a flood of fake AI‑generated reports, and by April most reports were AI‑written but still high quality.
Krueger’s call to action is simple: pick one open‑source project, identify its biggest install or runtime pain, and spend a weekend fixing it with the maintainers. Measure the gain, publish the results, and repeat. If AI can shave seconds off a pip install, it can also shave gigabytes off RAM use and hours off admin toil. The next wave of tech progress will be judged not by the number of shiny apps, but by how many critical tools run on a $5 server for a nonprofit of twenty people.
Sources: Hacker News, https://jordankrueger.com/blog/better-open-source-not-more-apps/