Devtools must be open source
Points and comments are a snapshot, not live.
Coding agents make personalizing open source software trivial, eliminating the need for plugin systems.
The author argues that LLM-powered agents fundamentally change the economics of personalizing software. Prompts like "download source, modify, and set up nightly cron to rebase upstream changes" let agents handle both initial customization and ongoing maintenance. This eliminates the need for config files or plugin systems. The author demonstrates by integrating a personal tool into their open-source agent Shelley with a single prompt. Closed-source agents like Claude Code lack this capability.
The implication is that end-user software must be open source to remain viable, as the cost of tailoring software to individual needs has effectively dropped to zero for those with access to source code and an agent.
What commenters are saying
Commenters broadly agree that LLMs make the open-source ideal more practical, but raise several concerns. Two main camps emerge: those who see AI as finally fulfilling the promise of user freedom to modify software, and those who warn of practical problems. Specific issues include: introducing subtle bugs or vulnerabilities without deep codebase knowledge; the burden of maintaining a private fork even with AI-assisted rebasing; and maintainers refusing AI-generated patches due to quality concerns, leaving users stuck with unmerged changes. One commenter notes that paying maintainers directly for PR review could solve the sustainability problem.